In 2005, a poet woke from the edge of sleep and heard a voice: “You have too many straight lines in your human world. We want to show you the way out of the violence of the line.” American poet CAConrad responded by abandoning left-aligned text entirely. A small act. A complete reorientation.

This week my friend George sent me a link to an exhibition of his late mother’s work. Anna Mendelssohn - poet, artist, revolutionary - kept hundreds of notebooks across a lifetime of difficult circumstances, published fifteen collections with almost no mainstream recognition, and left a vast archive, donated by her children to the University of Sussex after her death in 2009. Next month, her manuscripts are finally visible at the Fruitmarket Gallery in Edinburgh, exhibited alongside CAConrad’s work; two poets, across time and geography, brought together by a shared refusal of the straight line.

“The violence of the line” names something I have been circling since I left the academic world of language and literature and (naturally) became an accountant. The line. The straight line of argument, of hierarchy, of timeline, of the submission that must reach a recommendation by Tuesday. The line that insists on closure. The line that mistakes the answer for the destination and the destination for the point.

Historians have noticed this too.

Professor Daniel M. Knight is a philosophical, historical, and economic anthropologist at the University of St Andrews. His forthcoming book Against Polycrisis argues that the rush to declare our current convergence of crises as unprecedented is itself a problem. He calls it toxic presentism: the way our analytical gaze gets blurred by assumptions about linear time and historical progression, leaving us unable to learn from the many previous eras in which overlapping catastrophes arrived simultaneously and humanity, somehow, navigated them.

What strikes me about Knight’s argument is not that it is right, but how it is right. Walter Benjamin understood the same thing a century ago. His constellation method explicitly refused linear historical time; not out of nostalgia, and not as a licence for anything-goes thinking, but because the line suppresses what it cannot accommodate. The constellation allows moments across time to illuminate each other directly, sideways, without the violence of sequence imposing its meaning on them first. Benjamin’s method is Knight’s argument made methodological.

The line looks so reasonable, helpful even. Neutral. That’s its dark power.

In 1999, two researchers asked participants to watch a video of people passing basketballs and count the passes made by one team. Most participants, focused on their counting task, failed to notice when a man in a full gorilla suit walked into the middle of the screen, beat his chest, and walked off. The goal structure had made it structurally impossible to see anything else.

This is what the line does to human perception; participants were not stupid or careless. They were doing exactly what the goal structure required of them. That is precisely the point.

The line wields its power in every field. In academia, the line of argument - thesis, evidence, conclusion - is the entire apparatus of credibility. The peer review system, the citation economy, the conference circuit: a closure machine dressed as inquiry.

In the civil service, everything is distilled to a submission, a preferred option, a minister who needs an answer. The machinery of government was designed for resolution, not for the generative uncertainty that precedes it.

In the professions, the audit opinion, the legal judgment, the engineering sign-off are not merely outputs, they are the entire point of the formation. You cannot spend years being rewarded and credentialed for arriving at answers and then simply decide, in a Tuesday afternoon workshop, to state that you are happy to sit comfortably in the question instead.

This phenomenon even manifests in that most human of professions, coaching. An entire field ostensibly dedicated to creating conditions for people to think better has largely concluded that sitting in genuine silence, offering a Thinking Environment - a quality of attention so exquisite it allows the coachee’s own thinking to fully emerge, is not quite proper coaching. Not enough technique on display. Not enough line.

Systems thinkers know this too. The Built Environment Connective, a group of brilliant, committed people genuinely trying to do things differently in a sector that has resisted change for decades, produced a slide that captures the problem perfectly, without quite intending to. It shows the messy middle: overlapping circles, feedback loops, interconnected systems, human figures entering sideways. Everything the line cannot accommodate, drawn with honesty and care. And then, on the right, a clean green square. An outcome box. A 1, and a straight arrow pointing to it.

A spiral would be more honest - Jerome Bruner’s spiral learning that returns to the same territory at greater depth rather than arriving at a destination - but it gets straightened at the last moment into a line. Not from bad faith, but from the near-impossibility of leaving the outcome open in a room that was built for resolution.

Physicists have known this for fifty years. Ilya Prigogine won the Nobel Prize in 1977 for showing that complexity and apparent disorder are not the enemies of order; they are its only possible source. Novelty, adaptation, emergence: all of it lives at the edge of chaos, not beyond it, not resolved into a line.

My professional world is human infrastructure - the measurable, developable human capability layer that determines whether any major endeavour actually delivers. I am agnostic about sector, but engineering infrastructure, to which the United Kingdom has committed £725 billion of investment, currently offers the most vivid illustration of what happens when that layer is missing, and the most compelling evidence that everyone already knows it.

Between 2017 and 2026, the ICE, the IPA, NISTA, the Auditor General for Scotland, and the Scottish Parliament’s Public Audit Committee independently examined infrastructure delivery performance and each reached a version of the same conclusion: that human capability - the measurable, developable capacity of the people inside major programmes - is the critical and chronically unaddressed variable in delivery failure. Leadership churn. Loss of institutional memory. Culture that does not support value for money. Trust and collaboration named as essential, with no instrument identified for developing them.

Eight independent findings across a decade. Each ending in the same silence: no instrument for measuring or developing that capability was identified.

The line ran to the conclusion and stopped there. The gorilla walked off screen and nobody noted it on the scoresheet.

None of this is new; that is the data point.

In 2009, four Harvard Business School researchers published a paper that should have changed the conversation permanently. Goals Gone Wild argued that specific, challenging goals - the entire apparatus of performance management - produce systematic and predictable side effects: narrowed focus, unethical behaviour, distorted risk preferences, corrosion of organisational culture, reduced intrinsic motivation. They recommended that goal setting be treated not as an over-the-counter remedy but as prescription-strength medication requiring careful dosing and close supervision. They offered a warning label and called for a new generation of research.

Such is our attachment to goal-setting - the paper was largely ignored.

Sidney Harman said get me some poets as managers, poets are our original systems thinkers. Daniel Pink put it in a bestselling book. Margaret Heffernan has been making versions of the same argument for fifteen years. INSEAD Professor of Organisational Behaviour Gianpiero Petriglieri drew on Winnicott to show that what people need in uncertainty is not vision but holding; the capacity to contain distress, to help people make sense without rushing them to resolution, to be present rather than pointing endlessly at the horizon. Winnicott’s own insight, from the 1950s, was that holding is not what comforts people, it’s what makes them.

Jennifer Shahade, drawing analogies from chess in Thinking Sideways, says that too much linear focus on the goal means forgetting why you are on the path in the first place. The brain’s own optimal architecture, as neuroscience has been telling us for decades, is epistemic first, pragmatic second; curiosity before closure, uncertainty reduction before target pursuit. The moment you impose a specific challenging goal from outside you short-circuit the very process the brain uses to navigate complex environments well.

The only way out of stagnation is inner work. The compounding of developmental infrastructure over time - the grain of rice doubling on each square of the chessboard - cannot be achieved by reading a book or attending a training day.

Hou Yifan, one of the greatest chess players alive, understands this. Health and happiness, she says, is the 1; the baseline from which everything else compounds. The achievements, the victories, the outcomes are the zeroes that follow it. Without the 1, they are nothing. Place it first, and they multiply without limit.

Developmental infrastructure is the 1. Everything the reports have been asking for - the trust, the collaboration, the adaptive capability, the culture that holds under pressure - those are the zeroes. We have been trying to generate the zeroes without building the 1.

Right. Consistent. Well-evidenced. Widely read. The rooms haven’t changed.

The gap has never been between knowing and knowing more; it’s between knowing and being. The capacity to remain genuinely, epistemically curious, not performatively so, not as a technique deployed in a workshop, but as a sustained dispositional orientation toward what is not yet known. The 1 that makes everything else compound. It is not innate. It is not a personality type. It is not the exclusive property of poets or therapists or particularly enlightened leaders. It does not arrive through a workshop on psychological safety or an away day on systems thinking or a warning label on goal setting.

It is developable. Measurably, demonstrably developable. We know how. We have known for over a decade.

The question that remains is not whether the capability exists or whether it matters. Every report has confirmed that it does, every poet knows that it does. The question is whether the institutions that most need it can build the capacity to want it.

Anna Mendelssohn’s manuscripts sat in boxes at the University of Sussex for years after her death. Densely concentrated blocks of text and drawing, handwriting and image inseparable, resistant to every categorisation the straight line imposes. A critic once wrote to her: “Perhaps one image, one thing at a time?” He (we don’t know whether the critic was male or female, regardless the energy was masculine) meant it as guidance, but it was, without knowing it, a description of everything she was working against - the demand that the poem behave, that meaning be singular, that the constellation submit to the line.

The work that could not be contained is finally visible at the Fruitmarket in May, finally at the scale where you can see what it was doing all along: The poem that could not wait.

Every field has its version: the insight that arrived before the room was ready, that sat in boxes, that kept being passed over in favour of something more obedient, more linear, more suited to the closure the system demanded. Not because the insight was wrong. Because the line had no place for it.

🌹

Subscribe now

Further reading and references

This essay is dedicated to Anna Mendelssohn (1948–2009) - poet, artist, revolutionary, mother, who wrote under the name Grace Lake, who kept hundreds of notebooks in circumstances that would have silenced most people, who published fifteen collections with almost no institutional recognition, and whose defiant creed was: “I do not fake, I do not lie, I worship at the shrine of poetry.”

That she was my friend and WYSE-NGO colleague George O’Shaughnessy’s mother makes so much sense - I know him to be a quite remarkable character - passionate about co-operatives, grassroots organising and community empowerment. His mother would be extremely proud.

A note on Walter Benjamin and the constellation method

The risk with refusing to organise history as a straight line - from past to present to future - is that it looks like one of two things to a hostile reader.

The first is nostalgia: a sentimental preference for the past, a romantic backwards glance, a wish that things were as they used to be. Benjamin is not doing that. He is not saying the past was better. He is saying the past is alive - that certain moments speak directly to the present without needing to be arranged in sequence first.

The second is relativism: the position that because there is no objective linear progression, all moments are equally valid, all interpretations equally true, and there is no basis for judgment or knowledge at all. This is the charge levelled at anyone who refuses the Enlightenment timeline. Benjamin is not doing that either. He is not saying history is infinitely malleable or that meaning is arbitrary.

What he is doing is something more precise and more demanding: he is saying that the relationship between moments in history is not determined by their chronological position but by their structural resemblance - by what they illuminate in each other. The 1640s General Crisis speaks to the present not because it came before it, but because the configuration of forces rhymes. That is a claim about truth, not a retreat from it.

The same logic organises this reference list. It spans neuroscience, historiography, organisational behaviour, psychoanalysis, educational psychology, infrastructure policy, political science, poetry and chess - not because the argument is undisciplined, but because the same insight has arrived, independently, in each of these fields, and the fields have largely been unable to hear each other across disciplinary lines. The constellation holds what the timeline cannot, a challenge that shows up today in the fact that multiple professions keep arriving at the same “human capability is the issue” conclusion.

I. The Poem: Anna Mendelssohn and CAConrad

Mendelssohn (1948–2009), who also wrote as Grace Lake, kept hundreds of notebooks through circumstances that would have ended most writing lives and published fifteen poetry collections with almost no mainstream recognition.

In 1972 she was convicted in the longest criminal trial in British legal history at the time. Aged 24 Mendelssohn was one of the defendants brought to trial at the Old Bailey accused of conspiracy to cause explosions as part of The Angry Brigade. She maintained her innocence, insisting the conviction was unjust, and was sentenced to 10 years in Holloway women's prison. She was released on parole in 1976 and later adopted the name Grace Lake to escape the notoriety associated with the trial.

She became a prolific, highly experimental poet associated with the British Poetry Revival, with her work often reflecting on her experiences with incarceration, injustice, and her later academic life at Cambridge.

CAConrad. (2014). A Beautiful Marsupial Afternoon: New (Soma)tics. Wave Books.

CAConrad. (2019). While Standing in Line for Death. Wave Books.

Fruitmarket Gallery, Edinburgh. (2026). CAConrad & Anna Mendelssohn: the poem that could not wait. Exhibition, 1–24 May 2026. https://www.fruitmarket.co.uk/caconrad-anna-mendelssohn-the-poem-that-could-not-wait/

Clutterbuck, G. (2024). On Anna Mendelssohn. Annulet: A Journal of Poetics. https://annuletpoeticsjournal.com/George-Clutterbuck-On-Anna-Mendelssohn

Morrison, I. (curator). (2026). Exhibition notes: the poem that could not wait. Fruitmarket Gallery, Edinburgh.

Whitechapel Gallery. (2023). Speak, Poetess [Exhibition catalogue]. London: Whitechapel Gallery.

University of Sussex Special Collections. Anna Mendelssohn Papers. Archives Hub reference: 5603f8f1-ccfa-3637-aeb6-528e1c5f8965. https://archiveshub.jisc.ac.uk/search/archives/5603f8f1-ccfa-3637-aeb6-528e1c5f8965

Mendelssohn, A. (2020). I’m Working Here: The Collected Poems of Anna Mendelssohn (S. Crangle, Ed.). Shearsman Books.

Obituary: Anna Mendelssohn. (2009, December 15). The Guardian. https://www.theguardian.com/theguardian/2009/dec/15/anna-mendelssohn-obituary

II. Against the Line of History: Knight and Benjamin

Knight, D.M. (forthcoming, December 2026). Against Polycrisis. Stanford University Press.

Benjamin, W. (1999). The Arcades Project (H. Eiland & K. McLaughlin, Trans.). Harvard University Press. (Original work composed 1927–1940.)

Benjamin, W. (1968). Theses on the Philosophy of History. In H. Arendt (Ed.), Illuminations (H. Zohn, Trans., pp. 253–264). Schocken Books.

Lawrence, M., Homer-Dixon, T., Janzwood, S., Rockström, J., Renn, O., & Donges, J.F. (2024). Global polycrisis: the causal mechanisms of crisis entanglement. Global Sustainability, 7, e6. https://doi.org/10.1017/sus.2024.1

Tooze, A. (2022, October 28). Welcome to the world of the polycrisis. Financial Times.

III. The Gorilla: Inattentional Blindness and the Goal Structure

Simons, D.J., & Chabris, C.F. (1999). Gorillas in our midst: sustained inattentional blindness for dynamic events. Perception, 28(9), 1059–1074.

Ordóñez, L.D., Schweitzer, M.E., Galinsky, A.D., & Bazerman, M.H. (2009). Goals Gone Wild: The Systematic Side Effects of Over-Prescribing Goal Setting. Harvard Business School Working Paper 09-083.

Locke, E.A., & Latham, G.P. (2002). Building a practically useful theory of goal setting and task motivation: A 35-year odyssey. American Psychologist, 57(9), 705–717.

Staw, B.M., & Boettger, R.D. (1990). Task revision: A neglected form of work performance. Academy of Management Journal, 33(3), 534–559.

IV. The Rooms That Haven’t Changed: Institutional Evidence

Institution of Civil Engineers / Infrastructure Client Group. (2017). From Transactions to Enterprises. ICE. https://www.project13.info

Institution of Civil Engineers. (2020). A Systems Approach to Infrastructure Delivery (SAID). ICE. https://www.ice.org.uk/knowledge-and-resources/briefing-sheet/a-systems-approach-to-infrastructure-delivery

McNaughton, A. (2021, May). Why we must rethink major infrastructure projects. New Civil Engineer.

Institution of Civil Engineers / Infrastructure and Projects Authority. (2022). Transforming Infrastructure Performance Live. ICE. https://www.ice.org.uk/news-views-insights/latest-news/industry-must-embrace-change

Infrastructure and Projects Authority. (2025). Annual Report on Major Projects 2023–24. HM Government. https://www.gov.uk/government/publications/infrastructure-and-projects-authority-annual-report-2023-24

National Infrastructure and Service Transformation Authority (NISTA). (2025). Annual Report 2024–25. HM Government. https://www.gov.uk/government/publications/nista-annual-report-2024-2025

Audit Scotland. (2023). The 2022/23 audit of the Water Industry Commission for Scotland. Section 22 Report.

Audit Scotland. (2024). The 2023/24 audit of the Water Industry Commission for Scotland. Section 22 Report.

Scottish Parliament Public Audit Committee. (2025, May 16). Report: The 2022/23 and 2023/24 audits of the Water Industry Commission for Scotland. https://www.parliament.scot/about/news/news-listing/catalogue-of-failures-led-to-issues-at-scotlands-water-watchdog

Water Industry Commission for Scotland. (2025, November). Scottish Water’s Performance 2024–25. WICS. https://wics.scot/system/files/2025-11/Scottish-Water-Performance-Summary-2024-25.pdf

Institution of Civil Engineers. (2026). State of the Nation 2026. ICE.

Anderson, J., & Rainie, L. (2026). Building a Human Resilience Infrastructure for the AI Age. Imagining the Digital Future Center, Elon University. https://imaginingthedigitalfuture.org/wp-content/uploads/2026/03/ITDF-Human-Resilience-full-report-3.20.26.pdf

V. The Poets Were Always Right: Systems Thinking and the Creative Mind

Pink, D.H. (2005). A Whole New Mind: Why Right-Brainers Will Rule the Future. Riverhead Books.

Harman, S., quoted in Pink, D.H. (2005). A Whole New Mind. Riverhead Books.

Heffernan, M. (2011). Wilful Blindness: Why We Ignore the Obvious at Our Peril. Walker & Company.

Heffernan, M. (2023). Embracing Uncertainty: How to Thrive by Making Peace with the Unknown. Simon & Schuster.

Leslie, I. (2014). Curious: The Desire to Know and Why Your Future Depends on It. Basic Books.

Kay, J. (2010). Obliquity: Why Our Goals are Best Achieved Indirectly. Profile Books.

Bruner, J.S. (1960). The Process of Education. Harvard University Press.

Bruner’s spiral curriculum - the idea that learning returns to the same territory at increasing depth rather than resolving into a destination - is the developmental counterpart to the linear outcome model.

VI. Holding: The Container and the Messy Middle

Petriglieri, G. (2020, April 22). The Psychology Behind Effective Crisis Leadership. Harvard Business Review. https://hbr.org/2020/04/the-psychology-behind-effective-crisis-leadership

Winnicott, D.W. (1960). The theory of the parent-infant relationship. International Journal of Psycho-Analysis, 41, 585–595.

Winnicott, D.W. (1965). The Maturational Processes and the Facilitating Environment. Hogarth Press.

Kline, N. (1999). Time to Think: Listening to Ignite the Human Mind. Cassell.

McDowell, A. (2015). Building a Container. WYSE International working document.

Andrew McDowell’s framework for container-building - physical, behavioural, cognitive, emotional and transpersonal elements held together in service of genuine developmental work - was developed through UN-affiliated NGO WYSE International. Operating at the level of identity, purpose and psychological transformation, the twelve-day ILP is one of the most rigorous leadership development programmes in existence. WYSE does not describe the messy middle; it creates the conditions for people to inhabit it. The container will always be tested - that is how the group finds its boundaries. The question is whether the person holding it has the developmental capacity to remain present through the testing without reaching for the outcome box, the straight line. That capacity is precisely what WYSE builds, year after year, in facilitators and participants alike. https://www.wyse-ngo.org

Heifetz, R.A. (1994). Leadership Without Easy Answers. Harvard University Press.

Heifetz, R., Grashow, A., & Linsky, M. (2009). The Practice of Adaptive Leadership. Harvard Business Press.

VII. The Brain’s Own Architecture: Curiosity, Active Inference and Epistemic Value

Kaplan, R., & Friston, K.J. (2018). Planning and navigation as active inference. Biological Cybernetics, 112(4), 323–343. https://doi.org/10.1007/s00422-018-0753-2

Friston, K.J. (2010). The free-energy principle: a unified brain theory? Nature Reviews Neuroscience, 11(2), 127–138.

Gruber, M.J., Gelman, B.D., & Ranganath, C. (2014). States of curiosity modulate hippocampus-dependent learning via the dopaminergic circuit. Neuron, 84(2), 486–496. https://doi.org/10.1016/j.neuron.2014.08.060

Kang, M.J. et al. (2009). The wick in the candle of learning: Epistemic curiosity activates reward circuitry and enhances memory. Psychological Science, 20(8), 963–973.

Loewenstein, G. (1994). The psychology of curiosity: A review and reinterpretation. Psychological Bulletin, 116(1), 75–100.

Kashdan, T.B., & Fincham, F.D. (2004). Facilitating curiosity. In P.A. Linley & S. Joseph (Eds.), Positive Psychology in Practice (pp. 482–503). Wiley.

Kashdan, T.B. (2010). Curious? Discover the Missing Ingredient to a Fulfilling Life. Harper Perennial.

Rock, D. (2008). SCARF: A brain-based model for collaborating with and influencing others. NeuroLeadership Journal, 1, 44–52.

McGilchrist, I. (2009). The Master and His Emissary: The Divided Brain and the Making of the Western World. Yale University Press.

VIII. Thinking Sideways: Chess, Poker and the Compounding Baseline

Shahade, J. (2025). Thinking Sideways: Decision-Making Lessons from Chess. hodder press.

Kasparov, G. [@Kasparov63]. (2020, March). [Tweet on exponential growth and COVID-19]. Twitter. (referenced in Thinking Sideways).

Coren Mitchell, V. (2009). For Richer, For Poorer: A Love Affair with Poker. Canongate.

Dweck, C.S. (2006). Mindset: The New Psychology of Success. Random House.

Carol Dweck’s endorsement of Ruth Crick’s Learning Power framework: ‘Learning Power operationalises a growth mindset’ appears on the cover of Guy Claxton’s Building Learning Power (TLO, 2002).

IX. Configurational Thinking: Multiple Paths, No Single Lever

Mello, P.A. (2022). Incentives and constraints: a configurational account of European involvement in the anti-Daesh coalition. European Political Science Review, 14(2), 226–244. https://doi.org/10.1017/S1755773921000333

Ragin, C.C. (2008). Redesigning Social Inquiry: Fuzzy Sets and Beyond. University of Chicago Press.

The QCA methodology - which shows that outcomes emerge from configurations of conditions rather than single causes - offers the methodological counterpart to CLARA’s eight-dimensional Learning Power model. Neither admits reduction to a single variable.

X. The Developmental Infrastructure: CLARA and Learning Power

Deakin Crick, R. (2006). Learning Power in Practice: A Guide for Teachers. Paul Chapman Publishing / SAGE.

Deakin Crick, R., Broadfoot, P., & Claxton, G. (2004). Developing an effective lifelong learning inventory: the ELLI project. Assessment in Education, 11(3), 247–272.

Deakin Crick, R., et al. CLARA (Crick Learning for Resilient Agency). WILD Learning Sciences CIC. Validated across 100,000+ individuals in 190 organisations. https://www.wildlearning.co.uk

Claxton, G. (2002). Building Learning Power. TLO.

Keegan, R., & Lahey, L. (2009). Immunity to Change. Harvard Business School Press.

McGuire, J.B., & Rhodes, G. (2009). Transforming Your Leadership Culture. Centre for Creative Leadership / Jossey-Bass.

Fisher, D., & Torbert, W.R. (1991). Transforming Managerial Practice: Beyond the Achiever Stage. Research in Organizational Change and Development, 5, 143–173.

XI. The Long Thread: Evidence This Conversation Has Been Happening for Decades

Chamorro-Premuzic, T. (2014, August). Curiosity is as important as intelligence. Harvard Business Review. https://hbr.org/2014/08/curiosity-is-as-important-as-intelligence

Dyer, J., Gregersen, H., & Christensen, C.M. (2011). The Innovator’s DNA: Mastering the Five Skills of Disruptive Innovators. Harvard Business Review Press.

Collins, J. (2001). Good to Great: Why Some Companies Make the Leap and Others Don’t. HarperBusiness.

Senge, P.M. (1990). The Fifth Discipline: The Art and Practice of the Learning Organization. Doubleday.

Dilts, R. (1990). Changing Belief Systems with NLP. Meta Publications.

Robert Dilts’s Neurological Levels - environment, behaviour, capability, values/beliefs, identity, purpose - provide the architectural argument for why behaviour-level interventions fail without addressing identity and values. First formalised in 1990; still largely ignored by organisational development practice.

Pink, D.H. (2009). Drive: The Surprising Truth About What Motivates Us. Riverhead Books.

Tichy, N.M. (1997). The Leadership Engine: How Winning Companies Build Leaders at Every Level. HarperBusiness.

Marquardt, M.J. (2005). Leading with Questions: How Leaders Find the Right Solutions by Knowing What to Ask. Jossey-Bass.

Sullivan, W., & Rees, J. (2008). Clean Language: Revealing Metaphors and Opening Minds. Crown House Publishing.

Assagioli, R. (1973). The Act of Will. Viking Press.

Nir Eyal has written an article people need to read.

Published in Harvard Business Review this month, How Leaders Can Build a High-Agency Culture makes an important case: agency is not a personality trait, it is a learnable capability. The neuroscience behind it - Steven F. Maier and Martin Seligman’s work on the hope circuit, the research on mastery experiences and explanatory style - is compelling. The GE turnaround is a well chosen case-study. The four practices are coherent.

And yet - the verb in the title!

Build.

Carol Sanford, who spent decades thinking harder than almost anyone about what human development actually requires, had a name for this kind of framework. She called it other-directed development - the assumption, usually unconscious, that the job of a leader is to design experiences that produce the right outcomes in people. Audit their beliefs. Reframe their stories. Build systems that make the desired behaviour the default.

The contradiction is buried in that verb. You cannot build agency in another person. You cannot install it, engineer it, or activate it. The moment the experience is designed to produce it, you are producing something else - a more sophisticated form of compliance, a more elegant version of the behaviour the organisation already wanted.

The hope circuit - to use Eyal’s own image - isn’t something you wire up from the outside. It is already there. What is suppressing it is the question worth asking. The hand holding the string down. Not the scissors that might cut it free.

This is not a minor distinction. It determines whether what you build - forgive the verb - actually lasts.

***

The second thing the framework doesn’t quite reach is the measurement question, and this is where the real gap lies.

Every decade produces a new language for the thing organisations keep failing to build. Engagement. Psychological safety. Growth mindset. Learning culture. High agency culture. The language changes, but the gap persists.

One reason it persists is that without a validated instrument that tells you where adaptive capacity actually sits in a team - before and after, at individual and system level - you are operating on intuition masquerading as strategy. Telling people to audit their installed beliefs is useful advice. Knowing which specific dimensions of learning capability are constrained in this team, in this organisation, under these conditions, and being able to track whether they’re developing - that is a different order of rigour entirely.

In infrastructure procurement, where people and culture criteria are increasingly weighted and AI is making technical differentiation harder to sustain, we believe in our people is no longer sufficient. What evaluators, clients and regulators are beginning to ask for is proof. Not a narrative. Data.

***

The organisations that will actually build - grow - high-agency cultures are the ones that start from a different assumption entirely.

That people are the source of their own development, not the recipients of it. That the leader’s job is not to activate the hope circuit but to identify and remove what is suppressing it. That you measure the conditions that allow agency to emerge, track whether they’re changing, and hold yourselves accountable to the evidence.

Let go of the string.

The balloon knows what to do.

Subscribe now

Further reading

Martin Seligman began his career studying what breaks people. His 1967 experiments with Steven Maier - dogs who stopped trying to escape electric shocks because they had learned their actions made no difference - gave us the concept of learned helplessness, and with it a model of depression that reshaped clinical psychology.

What makes Seligman so interesting is what he did next. Dissatisfied with a profession organised entirely around pathology, he became the founding father of positive psychology - turning the field’s attention from what goes wrong to what allows people to flourish. Learned Optimism (1991) was the pivot. The question shifted from why do people collapse? to what allows them to keep going?

The final twist came in 2016, when Maier and Seligman revisited their original research using modern neuroscience and discovered their foundational theory was backwards. Passivity in response to sustained adversity is not learned; it’s the brain’s default. What has to be learned is the belief that effort matters. The hope circuit - the neural pathway that carries that belief - is subserved by the medial prefrontal cortex, which creates the conditions that shut down the passivity response. It does not come pre-installed - it develops through experience of genuine control.

The Hope Circuit (2018) is Seligman’s memoir of that whole journey. Highly recommended. Tons of free resources on The Positive Psychology website here.

Carol Sanford - The Regenerative Business (2017) and No More Feedback (2019). The most rigorous thinker on what human development actually requires. She would have had a great deal to say about this week’s HBR article on her brilliant podcast Business Second Opinion. Sadly we lost her in 2024.

Nir Eyal - How Leaders Can Build a High-Agency Culture, Harvard Business Review, March 2026. Worth reading.

On measurement - the question of how you actually read adaptive capacity in teams and organisations, rigorously and repeatedly over time, has been Professor Ruth Crick, founder of WILD Learning's work for twenty-six years. The instrument is called CLARA (Crick Learning for Resilient Agency).

Thirty years ago, a man walked into a primary school gymnasium in Dunblane, Scotland. It was a Wednesday morning in March.

I was a trainee accountant at Lothian Regional Council. I remember the call coming through to the office - not through a screen or a feed, but through a landline telephone, to my boss, the chief accountant for the Education department. That is how news travelled then; as a human voice, in an ordinary room, the weight of it landing at once.

There was no scrolling. No comments section. No way to be passively informed. I had been at Dunblane Hydro the weekend before. I didn’t know anyone involved; I didn’t need to. My immediate thought, and one I have held since, is what does it take to carry what they carry? What kind of world do we owe them?

What happened as a result of Dunblane is, I think, one of the most instructive things Britain has ever done.

⁂

Within weeks of the massacre, three women from the local area - Ann Pearston, Jacqueline Walsh and Rosemary Hunter - founded the Snowdrop Campaign. They were not politicians. They were not activists. By their own admission, they knew “nothing, really, about politics.” They were mothers who had lived in Dunblane for eighteen months, who had been part of that community, who understood with terrible clarity what had just happened and what it meant.

And, alongside bereaved parents including Mick North, they wanted the gun possession laws changed by the time the snowdrops bloomed the following spring.

The petition collected 705,000 signatures - in a pre-internet age, arriving in sacks of letters and cards. Their public profile brought both praise and death threats. Their PO Box was regularly closed because of bomb threats. They were told repeatedly that handguns couldn’t be banned because “pistol shooting is the fastest growing sport in the UK.” They kept going.

In less than a year of Dunblane, legislation banning handguns higher than .22 calibre was passed. Within twenty months private ownership of virtually all cartridge handguns was abolished. Great Britain looked at what had happened, looked at the tool that had made it possible, and made a collective decision about the relationship between dangerous capability and the society that holds it.

It has held. There has not been a school shooting in Britain since.

What made the difference, in part, was that two senior politicians - one Conservative, one Labour - walked into that gymnasium and saw what had happened with their own eyes. Lord Forsyth and George Robertson crossed party lines because the bodies of children made abstraction impossible. They could not unsee what they had seen. That is how political will sometimes works: not through argument, but through the removal of distance. They became, in the fullest sense, adults in the room - not because of their seniority or their politics, but because they had refused to let the decision be made from a safe remove.

The United States has made a different choice. Not once, but repeatedly, after Columbine and Sandy Hook and Uvalde and every name in between. The grief is not less. The horror is not less. The choice - and it is a choice, which is the uncomfortable part - has simply been different.

⁂

Last month, an eighteen-year-old killed eight people in Tumbler Ridge, British Columbia, and shot a twelve-year-old three times as she tried to lock the library door. The victims included five children aged twelve and thirteen, a teaching assistant, the shooter’s mother, and her eleven-year-old half-brother.

It has since emerged that in the months beforehand, she had been using OpenAI’s ChatGPT as a trusted confidant. A pseudo-therapist, a friend. Around a dozen OpenAI employees identified the interactions as indicating imminent risk of serious harm to others and recommended calling the police. Leadership declined. The account was banned. She opened a second account and continued.

A lawsuit filed in British Columbia alleges that ChatGPT was engineered to build psychological dependency - that the model was designed to assume best intentions, to echo back what users wanted to hear, to be whatever the person on the other side needed it to be. Not as a flaw, as a feature. Engagement over safety, at the level of the product’s founding logic.

OpenAI has since apologised. There will be inquiries and lawsuits. There will, eventually, no doubt, be regulation.

But let’s consider a different question: not the legal question, and not the regulatory one. The question underneath both of those.

What does it take to hold a powerful tool wisely?

⁂

The Dunblane answer - ban the tool - is not available for AI and it wouldn’t be the right answer if it were. The question is not whether to have the technology; it’s the same one Britain faced in 1996, reframed for a different kind of capability: what is the relationship between the power of the tool and the capacity of the people and institutions holding it?

In 1997, Britain decided that the gap between the destructive capability of a handgun and the average person’s capacity to hold that capability safely was too wide to bridge through licensing and training alone. The structural answer was to remove the tool from civilian hands entirely.

That calculation is not available with AI, but the underlying logic is worth keeping hold of: there is always a relationship between the power of a capability and the human capacity required to use it without harm. When the capability grows faster than the capacity, something breaks.

What broke in Tumbler Ridge was not, at its root, a technology failure. It was a human infrastructure failure - a young person in crisis, in a remote community, with no adequate mental health support, found a tool that was optimised to keep her engaged - to be whatever she needed it to be, to tell her what she wanted to hear, to hold her attention rather than her development. The technology did what it was designed to do. It just wasn’t designed with a vulnerable teenager in mind.

⁂

It is worth noting, without naïveté, that China is currently ahead of both the United States and the European Union in legislating for this specific problem. China’s proposed law on AI anthropomorphism - the use of AI to simulate human personality, thinking patterns and emotional connection - explicitly prohibits designing systems with the goal of replacing social interaction, controlling users’ psychology, or inducing addiction. It requires that providers possess “mental health protection, emotional boundary guidance, and dependency risk warning” capabilities. It mandates that users be reminded they are talking to an AI and creates specific protections for minors and the elderly.

The country that wants to win the AI race has looked at what is accumulating in courtrooms across the United States and decided to legislate against the design logic that made it possible. Tumbler Ridge is the most visible recent case - eight dead, a twelve-year-old with catastrophic brain injuries, OpenAI employees who knew and weren’t allowed to act. But it is not the only one.

Sewell Setzer III was fourteen when he died by suicide after forming an emotional attachment to a Character.AI chatbot modelled on a Game of Thrones character. Adam Raine was sixteen when ChatGPT, which his father described as transforming from a homework helper into a “suicide coach”, helped him plan his own death and offered to write his suicide note. Juliana Peralta was thirteen. There are at least six major cases pending, with more being filed. Character.AI and Google have already settled. The parents of these children stood before the US Senate Judiciary Committee in September 2025 and described, in devastating detail, what it looks like when a technology optimised for engagement meets a child in crisis. The EU has abstract provisions. The US is writing the rules case by case, in grief. China has looked at all of this and written a law.

One can hold China’s political context clearly and still recognise the precision of the diagnosis. There is an uncomfortable question sitting inside this comparison that shouldn’t be avoided. The line between a government surveilling its population and a government exercising a duty of care toward it is not always obvious. And at the furthest end of that same spectrum sits the autonomous weapon: AI deployed to kill, with no human shooter, no moment of human choice, no one to hold accountable. The removal of distance - the thing that made Forsyth and Robertson into adults in the room - becomes structurally impossible. There is no gymnasium to walk into. No bodies that makes abstraction unavoidable.

That line becomes harder to locate when the population has had its agency so systematically eroded - by engineered food, by addictive social media, by AI designed for dependency - that it is worth questioning whether the capacity for genuine autonomous choice is already compromised.

The liberal answer, that adults must be free to make their own decisions, assumes an adult whose decision-making capacity has not been deliberately undermined. When that assumption fails, someone has to be the adult in the room. China has decided, for its own reasons and in its own way, that this is a state function. The West has largely decided it isn’t. Tumbler Ridge is one data point in the case for reconsidering that position.

This is what the absence of governance architecture looks like. Not dramatic, just a gap where the decision should have been.

⁂

This week I watched 1,400 people comment on an Instagram post about quitting ChatGPT. The dominant note was passivity: “Me cancelling my subscription will do nothing.”

That response is itself a data point. The feeling of having no meaningful agency in relation to a technology that shapes your daily life is not a personality trait. It is a learned condition. Platforms designed to cultivate dependency rather than capacity produce exactly this: people who have lost the felt sense that their choices connect to outcomes.

This is not accidental. It is the product. And it is worth asking - as we ask it of food systems, as we ask it of social media - when a technology is deliberately engineered to produce passivity, at what point does calling it a failure of personal responsibility become a way of protecting the people who built it?

Ann Pearston and Jacqueline Walsh and Rosemary Hunter knew nothing about politics. They had received death threats. Their PO Box was being bombed. They were told the fastest growing sport in the UK depended on the thing they were trying to ban.

They kept going anyway because they refused to believe their choices would have no impact.

⁂

We are at the beginning of a period in which extraordinarily powerful tools are being placed in the hands of individuals, organisations, and nations whose capacity to hold them wisely varies enormously - and is almost never measured.

The question for the AI era is whether we are capable of the same quality of collective decision-making that Britain managed in 1996. Whether we can look at what is happening, clearly, without flinching, and ask not just what the technology can do, but what we need to become in order to hold it wisely.

That is a question about human capability. It is also, if we are serious about it, a question about infrastructure.

⁂

Thirty years ago, three women in Dunblane who knew nothing about politics decided that the murder of sixteen children and their teacher was not the end of the story. They collected 705,000 signatures in sacks of letters and cards, endured death threats, and kept going until Parliament acted. Two politicians walked into a gymnasium and refused to look away. Within eighteen months, handguns were banned. There has not been a school shooting in Britain since.

Since Columbine in 1999, at least 218 children and educators have been killed in school shootings in the United States. The US has had 57 times as many school shootings as all other major industrialised nations combined. The choice made after Dunblane has held; the choice not made has compounded.

Now a different kind of tool is placing a different kind of pressure on a different generation of children. AI companies that claim to put safety at the heart of what they do will have that claim tested; in the courts, in the regulatory hearings, in the grief of communities. The question is whether we wait for the evidence to accumulate, or whether we build the governance architecture and the human infrastructure - the capability to use tools wisely - before the next preventable tragedy makes it unavoidable.

We can make choices too, but only if we haven’t lost the Will to make them.

Sources and further reading

Dunblane: How Britain Banned Handguns. BBC Scotland / BBC Two, broadcast 10 and 12 March 2026. Produced by IWC Media. Features Ann Pearston, Jacqueline Walsh and Rosemary Hunter (founders of the Snowdrop Campaign); bereaved parents Mick North and Kenny and Pamela Ross; PE teacher Eileen Harrild; Tony Blair; Lord Forsyth; Lord Robertson; and Alastair Campbell. Available on BBC iPlayer.

‘Our children paid the ultimate price’ - How the Dunblane school shooting changed Britain. PA Media / Yahoo News, 8 March 2026.

Campaigners faced ‘death threats’ over fight for handgun ban after Dunblane. LBC, March 2026.

Family of Tumbler Ridge shooting victim sues OpenAI. CBC News, 9 March 2026.

Mother of wounded Maya Gebala sues OpenAI over mass shooting in Tumbler Ridge, B.C. The Canadian Press, 10 March 2026.

Families of Tumbler Ridge victims pursuing lawsuits against AI companies could face long journey. The Globe and Mail, March 2026.

Parents of 16-year-old Adam Raine sue OpenAI, claiming ChatGPT advised on his suicide. CNN Business, 26 August 2025.

Google and Character.AI agree to settle lawsuits over teen suicides linked to AI chatbots. CNN Business / Fortune, January 2026.

China’s Approach to AI Anthropomorphism. Luiza Jarovsky, PhD. Luiza’s Newsletter, 13 January 2026. luizasnewsletter.com Find Luiza on Substack here

Interim Measures for the Administration of Humanized Interactive Services Based on AI. Cyberspace Administration of China, December 2025.

Share

Since 2023, some of the smartest men in the world have been standing in the kitchen, staring at the refrigerator. The ingredient isn't in there, sweetheart. It's in the cupboard. It's always been in the cupboard.

They think it’s data.

Large Language Models are refrigeration. Essential, world-changing but not ultimately where the real money will be made. The Coca-Cola companies, the ones with one extra ingredient that produces markedly different outputs from the same inputs - those are the ones to watch.

An elegant analogy by Chamath Palihapitiya, the man who scaled Facebook from 50 million to 700 million users and now manages $2 billion of his own capital.

He's right, he just hasn't opened the right cupboard.

To be fair to him, he has been looking and since the analogy started circulating the kitchen has got busy. A principal technical product manager at Mayo Clinic concluded the secret ingredient is proprietary data: Netflix viewing history, Tesla driving miles, Uber GPS trails. An advertising strategist argued it's the ability to generate what he called narrative violations at scale - creative work so context-specific and surprising it couldn't exist without AI. An entrepreneur spotted that the boutique phase is coming, and the winners will be the ones who already have the trust. A VC firm published seventeen hundred words on Coca-Cola, Walmart and supermarkets without once asking who’s actually makes the food.

Smart men, every one of them, but same blind spot, every time.

The extra ingredient they keep reaching for is assumed to be technological. Data, features, distribution, architecture, product. They are looking for it in the refrigerator, on the shelves, in the supply chain. It has not occurred to any of them that it might be standing right there in the kitchen. Has been, for rather a long time and knows exactly where everything is.

The blind spot has a history.

Facebook, the company Chamath Palihapitiya helped scale to planetary reach, began as FaceMash, a website built by a Harvard undergraduate to rate the attractiveness of women on campus. You uploaded two photos, clicked the hotter one. The algorithm learned what made people react.

That’s it, that’s the founding impulse. Reduce a woman to a data point. Extract a reaction. Measure the engagement. Scale it to two billion people.

And now the man who ran that growth engine is telling us that the secret ingredient in the intelligence economy is data.

Data matters, but not that kind.

Palihapitiya spent his career measuring what humans do when you make them reactive. Scroll, react, share. Signal for human receipt, perform connection rather than build it. The attention economy didn’t develop human relational capacity, it monetised the hunger for it. Two billion people, exquisitely measured, not one of them more capable of learning, adapting or thinking together than they were before they signed up.

The ingredient he’s missing - the one nobody in this conversation has named - is what humans become when they have agency.

Reflect, question, adapt, learn. Collaborate under pressure. Make meaning together when the situation has no precedent and the model has no answer. That is not a dataset. It cannot be scraped, purchased or reverse-engineered. It takes, as it happens, about twenty-six years to build properly.

Jensen Huang gave us a five-layer cake: energy, chips, cloud, models, applications. I wrote about the missing ingredient in that framework last week - not a sixth layer sitting on top, but something more like baking powder. Invisible in the finished product, but present through every layer. Human infrastructure. Entirely responsible for whether you end up with something edible or an expensive, flat mess.

Chamath gives us Coca-Cola, which is, if anything, a better domestic metaphor. Because the secret formula for Coca-Cola was not data, or a feature or a distribution strategy or a proprietary algorithm. It was a recipe, developed by a pharmacist, refined by hand, held in trust, not replicated because it could not be replicated. It was knowledge that lived in people, passed between people, developed over time by people who understood what they were making and why it mattered.

The AI companies looking for their Coca-Cola moment are asking: what’s the extra ingredient?

The answer has been sitting in plain sight, in the domestic economy, in every kitchen and classroom and community that has ever actually functioned, for rather longer than the venture capital industry has existed.

It’s the capacity to learn. To adapt. To work alongside people who are different from you, under conditions you didn’t anticipate, toward goals that keep shifting. To keep asking better questions when there are no easy answers left.

That capacity - not a soft skill, not a culture initiative, not a vibe - has been quietly measurable for twenty-six years. Professor Ruth Crick, a former music teacher who became one of the world’s leading researchers in learning science, has spent that time building the instrument. It’s been tested on more than 100,000 individuals across 190 organisations. It predicts resilience, performance, adaptive expertise and purposeful behaviour under conditions of complexity. The Scottish Qualifications Authority has accredited the framework from Level 1 to Level 7.

Nobody else has it because nobody else built it. You cannot buy it from a vendor. You cannot generate it with a model. You cannot, it turns out, find it by staring at the refrigerator.

The 51% of organisations that McKinsey reports have experienced negative outcomes from AI adoption are not suffering from a data problem. They are not victims of insufficient model sophistication or inadequate product strategy. They deployed intelligence into systems whose people could not metabolise what it gave them. Teams adopted AI individually and the collective intelligence of the organisation fell, because people stopped thinking together. They outsourced cognition to the machine and the capability stayed with the machine, not with the people.

We have been here before. FaceMash rated women. The attention economy measured humans. Follow the same logic into AI and we don’t get Coca-Cola. We get something worse: humans optimised for compliance, capability quietly atrophying, cognition outsourced to the machine. Nodding dogs. If whoever tells the best story wins, right now men are telling a very compelling data story and we are all, collectively, nodding along.

And the story may be running out of road. Sinead Bovell put it better than most: we may be at the end of the social media era. If AI breaks the psychology of signalling for human receipt - if you can no longer tell whether the 14 million views are humans or bots - then the entire value proposition of the attention economy quietly cracks. We performed connection for two decades because humans were watching. If they’re not watching, the performance has no audience.

Which means the organisations that invested in actual human relational capacity - rather than its performance - are about to discover they have a structural advantage that nobody priced in. Not because they were virtuous. Because they built something real while everyone else was building something that required an audience to function.

The ingredient in the cupboard is the alternative to that story.

The refrigerator is still running. The ingredient is in the cupboard. It has always been in the cupboard, sweetheart.

References and further reading

’s refrigeration analogy has been circulating since December 2023, originally shared via Vala Afshar on X. The most thorough written treatment is Bedrock Group’s August 2024 piece, The Coca-Colas of AI.

Sinead Bovell on the end of the social media era - the observation about signalling for human receipt - comes from this conversation which is worth listening to in full: The Social Media Era Is Ending: AI, Voice and the Future of Attention

The foundational academic paper behind the Learning Power framework and the argument that adaptive capacity is measurable and developable: Ruth Deakin Crick, Developing Resilient Agency in Learning: The Internal Structure of Learning Power, 2015: tandfonline.com

The Missing Ingredient — the Jensen Huang / baking powder piece this one builds on is here

In 2001, if you asked a reasonably well-informed person to name the two most important search engines, they would probably have said Google and Yahoo. And they wouldn’t have been wrong. Both were credible. Both were growing. Both were competing for the same users, the same advertisers, the same slice of an internet that was still figuring out what it was.

By 2010, the question would have sounded almost confused. Google and Yahoo weren’t competing anymore, not really. They had become different things. Google had become infrastructure: the underlying layer through which people navigated the web, the default assumption baked into browsers and operating systems and daily habit. Yahoo had become media: a portal, a homepage, a collection of content properties and advertising revenue. Same surface in 2001. Completely different asset class by 2010.

Nobody announced the category split. Nobody scheduled it. It happened through accumulated choices; what each company prioritised, what it refused, what it was willing to sacrifice in the short term for something it believed about the long term. The divergence was clear only in retrospect, and only to people who knew what they were looking for.

The category divergence wasn’t just in the strategy - it was already in the architecture. While other search engines indexed pages by keywords, Google’s PageRank measured a page’s importance by how many other pages linked to it, and how important those pages were in turn. It was a different theory of what search fundamentally was, and the bigger the web grew, the better it got, because it had more signal to work with

I’ve been thinking about this a lot since I read the memo that Dario Amodei sent to Anthropic’s employees.

The background: the US Department of Defense approached Anthropic about an AI contract. Anthropic held its red lines: there were use cases it wouldn’t support, conditions it wouldn’t agree to. The contract went to OpenAI instead. Sam Altman framed Anthropic’s position as unreasonable, inflexible, bad faith. Dario pushed back directly: the real issue, he wrote, is that Anthropic hasn’t donated to Trump, hasn’t offered dictator-style praise, has supported AI regulation, and has told the truth about things - including job displacement -that the current administration finds those truths inconvenient.

The commentary around this has mostly treated it as a story about political courage, or corporate ethics, or the inevitable tension between principle and revenue. Those readings aren’t wrong, but I think they’re missing the more interesting thing.

What Dario is actually describing is a strategic fork, and the choices being made in this window, by these two companies, will determine which category each of them ends up in.

Here is what I think is true, and what I haven’t seen anyone else say yet: we are watching the AI industry do exactly what the search industry did in the early 2000s. Right now, OpenAI and Anthropic look like competitors. They make similar products. They recruit from the same talent pool. They’re fighting for the same enterprise contracts. The commentary treats them as horses in the same race.

But the choices they’re making aren’t just strategic variations on a shared theme. They are bets on what the technology fundamentally is.

OpenAI is making a media bet. Distribution, reach, consumer dominance, content, commerce, political access. The logic is familiar from every platform war: get big, get sticky, get into the operating system of daily life. Altman’s moves - the Trump relationship, the DoD contract, the consumer expansion, the aggressive positioning - are all consistent with a company that wants to be the destination. The place people go, the thing they open.

Anthropic is making an infrastructure bet. Slower, more constrained, more expensive, defined as much by what it won’t do as what it will. The Constitutional AI framework, the safety research, the enterprise focus, the refusal to take certain contracts; these aren’t just values statements. It’s an architectural choice - a different theory of what an LLM is and what it’s for. They are the structural conditions that make trust possible at institutional scale. Anthropic is trying to become the layer underneath, the thing that’s simply there, that institutions rely on the way they rely on the power grid or the internet protocol.

Infrastructure and media are not just different business models. They are different asset classes, with different relationships to time, trust, and risk. Media is won in cycles, infrastructure compounds over decades. Media depends on attention, infrastructure depends on reliability. Media companies can reinvent themselves. Infrastructure, once embedded, is almost impossible to displace.

The analogy isn’t perfect because analogies never are. Google didn’t have to navigate a geopolitical moment quite like this one, and the stakes in AI are orders of magnitude higher than search. But the underlying dynamic is the same: two companies that look like competitors are actually resolving into two different things, and most observers won’t notice until it’s already happened.

What makes the current moment particularly sharp is that the divergence is being accelerated by external pressure. The Trump administration’s enthusiasm for OpenAI and hostility toward Anthropic isn’t just political weather, it’s a forcing function. It’s making each company’s fundamental bet visible earlier than it might otherwise have been. Altman’s pragmatism is the pragmatism of a media company: access and distribution matter more than any particular principle. Amodei’s refusals are the refusals of an infrastructure builder: the integrity of the layer is the product and compromising it for a contract is not a trade worth making.

Dario Amodei called out the gaslighting directly in his memo, which was itself an unusual move: CEOs of companies at that stage don’t usually send employees a detailed account of political manoeuvring and competitive positioning. But it makes sense if you understand what he’s actually protecting. It’s not just Anthropic’s reputation, it’s the structural condition - independence, integrity, credibility - that makes the infrastructure bet viable at all.

The search wars ended the way most technology category splits end: slowly, then suddenly, and then it seemed obvious all along.

But here’s what the analogy misses, and what I’ve been sitting with since a conversation this week with a friend who runs cyber security contracts for the Home Office. She’s one of the smartest people I know and her work is actually impacted by the Department for War’s stance, and yet she has never used an LLM. Not because she can’t, but because she doesn’t trust what happens to information she puts into it. She is not an outlier.

The commentators and the industry analysts are asking which horse will win the race. Rutger Bregman just published a Guardian op-ed calling for a mass boycott of ChatGPT, and more than a million people have already cancelled their subscriptions in response. The Instagram comments beneath it were running nine-to-one in favour of not using AI at all. “Resistance is not using it.” The race the commentators are watching is not the race most of the world is in.

Which makes the category split even more consequential than the Google/Yahoo story suggests. Because the third group - the friend running cyber contracts for the Home Office, the Instagram commenters, the hundreds of millions of people who are not tech workers, not early adopters, not ideologically invested in Silicon Valley’s optimism - that group is not waiting to see who wins. They are waiting to see whether they can trust any of this at all.

Infrastructure wins that group. Media doesn’t.

The company that becomes the layer underneath - reliable, principled, present without demanding to be noticed - is the one that has a chance at earning the trust of people who never signed up for the race in the first place. That’s a much bigger market than the one everyone is currently arguing about.

The AI category split is happening now, in real time. Most commentators are still asking which horse will win the race. Most of the world isn’t watching the race at all. The most interesting question - the one that will look obvious in ten years - is which company understood that.

What do you think - are you seeing the same divergence? I’d love to know.

In the summer of 1914, Europe had some of the most capable leaders it had ever produced. Educated, experienced, advised by generals who had studied every previous war. The telegraph was modern technology. The railway network was the infrastructure of the age. The plans were meticulous.

Nobody decided to destroy a generation; the plans ran the people.

Barbara Tuchman’s The Guns of August - the Pulitzer Prize-winning account of how the First World War began - is not a book about stupidity or malice. It is a book about what happens when systems exceed the relational capacity of the people inside them. When the machinery of coordinated action becomes more sophisticated than the human ability to pause it, question it, hold the other side as legitimate, keep the back-channel open.

It has been on my mind lately because of two things that happened on the same day.

The first: Professor Ruth Crick, the researcher whose Learning Power framework underpins the work I do, sent a reflection from her holidays on a beach in Malaysia. She had been reading through two years of strategic documents. Her observation was quiet and precise: if WILD became a field embedded in organisational AI tools, it could potentially introduce a very different orientation; learning as an ongoing relational process rather than a technical optimisation.

The second: watching a podcast - Daniela Amodei, Co-Founder and President of Anthropic, sat in the library of Anthropic’s San Francisco headquarters and recommended The Guns of August to a podcast audience. Not casually, but as the book she’d reach for right now, in the context of a conversation about AI governance, the pace of model development, and what we’re about to ask of the humans steering it.

Two women. Different hemispheres, different disciplines, different vantage points on the same question. And the question, when you strip it back, is the one Tuchman asked in 1962 and Kennedy heard clearly enough to act on during the Cuban Missile Crisis: what happens when the system gets faster than the people running it?

The biological argument

The Chilean biologist Humberto Maturana spent his career studying cognition and emotion, and he arrived at a conclusion that is either obvious or radical depending on where you’re standing: love is the only emotion that expands intelligence.

Not love as sentiment. Love as the capacity to hold the other as legitimate; as a genuine other with their own reality, their own perspective, their own claim on the situation. What Maturana calls the radical acceptance of the other.

Fear, by contrast, contracts intelligence. When we are threatened, the cognitive architecture narrows. We process faster but less well. Pattern recognition accelerates; nuance disappears. We see threats where there are ambiguities, enemies where there are merely differences. We stop learning and start defending.

This is not metaphor. It is neuroscience. The vagus nerve, the architecture of the parasympathetic nervous system, the embodied cognition that Stephen Porges documents in the Polyvagal Theory - the ventral vagal system as the neurological substrate of social engagement, the physiological precondition for relational thinking itself. It expands under conditions of safety, curiosity and belonging and contracts under threat.

The leaders of 1914 were not stupid, they were afraid. And fear-contracted cognition, operating inside locked institutional systems, produced outcomes that no individual would have chosen and that no individual had the relational capacity to stop.

Kennedy read “The Guns of August” during the Cuban Missile Crisis and kept the back-channel open. He refused to let the plan run the people. That refusal - choosing to expand the relational field at the moment fear was most aggressively contracting it - is the only decision that matters in retrospect.

The muscle Kennedy exercised in October 1962 is the same muscle organisations are being asked to exercise right now, in conditions of AI-induced uncertainty, at a speed that the institutional architecture was not designed for.

It is not a technical muscle, it is a relational one, and we have no governance architecture for it.

The measurement error

We have been making a civilisational measurement error for several centuries. We decided - not formally, not consciously, but structurally and persistently - that quantifiable intelligence was real intelligence, and that relational intelligence was something softer. Secondary. Feminine. Suspect.

The intelligence that gets you into the right university. Through the right pipeline. Onto the stage. The intelligence measured in grades, benchmarks, processing speed. That intelligence, the one we built institutions to produce and reward.

The other kind - the capacity to hold complexity, to maintain curiosity under pressure, to collaborate across difference, to stay in the question rather than retreating to the answer - that intelligence we acknowledged in principle and systematically undervalued in practice. It had no measurement instrument. It had no governance architecture. It was, as the economist would say, a public good that no one owned.

This was not an innocent oversight. It had an architecture, it had victims.

Alfred Binet designed what became the Stanford-Binet intelligence test in 1905 for a specific and humane purpose: to identify children in the Paris school system who needed additional support. He was explicit, in his own writing, that the test measured a narrow slice of cognitive function under particular conditions. It was not, he insisted, a measure of fixed innate intelligence. It was a diagnostic tool for targeted help.

Lewis Terman at Stanford took the test, renamed it, reframed it as a measure of hereditary intelligence, and built an entire scientific apparatus around the claim that IQ was the truest measure of human worth. The intelligence it measured - abstract reasoning, pattern recognition, verbal facility - happened to be the intelligence that formal Western education produced and rewarded. The intelligences it could not see - relational, embodied, creative, contextual - were reclassified as absence. As deficiency.

The consequences were not merely philosophical. Eugenics programmes across the United States used the Stanford-Binet as scientific cover for the forced sterilisation of more than 60,000 people deemed insufficiently intelligent. The targets were disproportionately women, immigrants, the poor, and people of colour; the people whose intelligence expressed itself in precisely the forms the test was not designed to detect. The Nazi eugenics programme, when building its own architecture, explicitly cited American practice as its model.

Binet died in 1911. He did not live to see what was done with his instrument. But he had already warned against it, which makes the history not a story of scientific error but of deliberate misappropriation. The measurement error was chosen, the hierarchy was enforced. And the people who paid for it were the ones whose intelligence had no official name.

The attempts to correct this have been persistent, and persistently recaptured.

Every generation since has produced a man with a new metric. Emotional intelligence was the most prominent - Daniel Goleman’s EQ arriving in 1995 with genuine insight at its core: that emotional attunement, empathy, and self-regulation matter as much as abstract reasoning, perhaps more. The insight was real, what happened to it was not. EQ was rapidly absorbed into the same reductive apparatus - competency frameworks, five-point scales, certification programmes, corporate training days. The relational was annexed into the quantifiable. The challenge was neutralised. What had briefly threatened the hierarchy became another column in the spreadsheet, another line on the CV, another thing you could score higher on than the person next to you.

The pattern repeated because the method was always the same: notice the unmeasured thing, measure it badly, commodify it, and in doing so confirm rather than challenge the original hierarchy. The question was never seriously asked - not in the boardroom, not in the policy paper, not in the business school - whether the framework itself was wrong. Whether intelligence might not be a spectrum from low to high but something altogether more plural, more embodied, more irreducibly relational than any single instrument could capture.

In his book Intelligence in the Flesh, Guy Claxton asked that question. Not by adding another dimension to the cognitive scorecard but by going to the body entirely - the three brains, the gut intelligence, the nervous system as the substrate of knowing. Stephen Porges provided the physiological mechanism: the ventral vagal system as the neurological precondition for social engagement, for the kind of relational thinking that no IQ test and no EQ framework had ever been designed to see. And Humberto Maturana provided the ecological condition: that intelligence of this kind does not exist in individuals. It exists between them. It expands or contracts depending on the relational field; on whether the other is held as legitimate or as threat.

These are not softer versions of the original hierarchy, they are a different account of what intelligence is and where it lives. That account has been available for decades. It has simply had no governance architecture, no measurement instrument, and no institutional home powerful enough to make it stick.

AI is now correcting this valuation error structurally, and faster than anyone anticipated.

Jensen Huang put it plainly. When asked who the most intelligent person he had ever met was, he declined to answer, because the question revealed the problem. Technical problem-solving, he observed, is what AI solves easiest. Software programming was supposed to be the ultimate marker of intelligence. It was the first thing to go. So the definition of intelligence itself needs to change.

Daniela Amodei, when asked what parents should tell their children to study, said: it doesn’t matter what they major in. It matters how they treat the people they work with.

That is Maturana without the vocabulary. It is the relational condition named as the thing that matters, by the president of the most consequential AI company in the world, in a library full of books, in a conversation about what comes next.

She doesn’t have the measurement instrument. She knows she doesn’t have it, the judgment question in that podcast got a values statement rather than an answer. But she is asking the right question, which is where everything starts.

The plan running the people

The specific danger of the current moment is not that AI will become malevolent, it is that AI will become the plan.

That organisations will deploy AI systems with extraordinary capability, and that the humans inside those organisations will gradually, not through any single decision, not through malice, not through stupidity, stop being the agents of their own thinking. Will outsource judgment to the tool. Will optimise for the tool’s metrics. Will, in Tuchman’s language, find themselves carried forward by machinery they once set in motion and no longer know how to stop.

This is not a futurist scenario. It is already the observable pattern in every technology adoption curve. The question is not whether it will happen. The question is whether the organisations of the 2020s will build the relational infrastructure to prevent it before a crisis forces the architecture, or after.

Sener Cem Irmak, founder of Koru Impact, identified the exact precedent. Risk management spent decades structurally orphaned - acknowledged in principle, owned by nobody - until 2008 forced it into governance architecture that boards could no longer ignore. Human infrastructure is at that inflection point now. The question is whether we build the architecture before the crisis, or after.

History suggests most will wait. But the organisations that don’t will have a structural advantage that compounds quietly over years, in exactly the way that human infrastructure does, and technical infrastructure does not.

What the field condition changes

Ruth’s observation from the beach is precise about what is actually at stake in the AI tools conversation.

Most AI tools, as currently designed, optimise for resolution. Get the user to the answer faster. Reduce friction. Complete the task. The logic, when deployed at organisational scale, doesn’t merely skip learning - it can quietly erode the disposition to learn. The curiosity that sustains inquiry. The tolerance for productive struggle. The relational capacity to think with others rather than through a tool.

A field condition is different. It is not a feature added to a system. It is the ambient orientation that shapes how every interaction unfolds. The difference between an environment that treats interactions as transactional - what is the output? - and one that treats them as developmental - what is being built in the person doing this?

WILD’s Learning Power framework - eight empirically validated dimensions including curiosity, sense-making, collaboration, mindful agency, and openness to learning - measures not skills, not knowledge, but the adaptive capacity that allows people to keep learning as conditions change. The disposition to engage with complexity rather than retreat from it. The relational intelligence to work across difference.

These dimensions are what Kennedy drew on in October 1962. They are what Daniela Amodei describes when she says the thing AI cannot do is interact with humans the way humans do. They are what Maturana means by the relational condition that expands intelligence.

And they are currently unmeasured, unowned, and ungoverned in every organisation deploying AI at scale.

The constitutional question

Anthropic’s Constitutional AI is - as Daniela described it in the library - an attempt to give Claude a framework for thinking about ethics, rather than a reward-and-punishment system for individual answers. Not ‘this answer is good, this answer is bad,’ but: here is the system of values, the relational orientation, that should govern how you approach any question.

What Ruth names is the equivalent constitutional question at the human level: what orientation does a tool embed in the people using it? Not what does it produce, but what does it cultivate? What kind of learner, what kind of thinker, what kind of collaborator does using this tool, over time, at scale, tend to produce?

That is a question the tools industry does not ask. It is not asked because the answer is not technically measurable with existing instruments. And what is not measurable cannot be governed, funded, or held accountable.

The measurement instrument exists. Ruth Crick’s Learning Power research spans twenty-five years, structural equation modelling, peer-reviewed publications, and a validated diagnostic - CLARA, the Crick Learning for Resilient Agency - that measures the orientations and relational conditions that predict adaptive capacity over time. It is not a training assessment. It is a longitudinal instrument for the thing that currently has no owner.

You cannot govern what you cannot measure. And the relational infrastructure of human intelligence - the thing that keeps the plans from running the people - has been unmeasurable until now.

Two women, one question

Ruth on a beach in Malaysia. Daniela in a library in San Francisco. They are not in the same conversation yet.

But they are asking the same question, from the same intuition, with the same clarity about what is missing. One has spent thirty years building the measurement instrument for what the other has named as the thing that matters most.

The Guns of August is a book about what happens in the absence of that instrument. When capable, well-intentioned people operate in systems that have exceeded their relational capacity, with no architecture for measuring or governing the thing that would have changed the outcome.

We are not in 1914. The conditions are different and the risks are different, but the structural logic is the same: technology has deployed faster than the human infrastructure required to govern it wisely. The plans are running the people. And the relational capacity that would allow us to pause, question, hold the other as legitimate, keep the back-channel open - that capacity has no owner.

The question is not whether we need governance architecture for human relational intelligence. We do. The question is whether we build it before the crisis that forces it, or after.

Kennedy read the book. He kept the back-channel open. Thirteen days later, the world did not end.

What he had, in the moment that mattered, was not better technology. It was a relational capacity that had been deliberately developed, that could hold fear without contracting into it, that could see the other side as legitimate even under maximum threat.

That is what WILD measures. That is what no one currently owns. That is what Ruth named from a beach in Malaysia, and Daniela reached for in a library in San Francisco, in documents that landed on the same desk on the same day - from opposite sides of the world.

Further Reading

The following works form the intellectual substrate of this piece. They are worth reading in this order - each one opens the next.

On the historical argument

Barbara W. Tuchman, , The Guns of August (1962). The foundational text. Tuchman traces the five weeks before and after the outbreak of the First World War, showing how capable, well-intentioned leaders were carried forward by institutional machinery that exceeded their relational capacity to govern it. Kennedy kept a copy on his desk during the Cuban Missile Crisis.

On the biology of intelligence

Humberto Maturana & Francisco Varela, , The Tree of Knowledge (1987). The biological argument for love as the condition of expanded intelligence. Maturana’s concept of autopoiesis - the self-creating, self-maintaining nature of living systems - is the foundation for understanding why relational conditions matter as much as cognitive ones.

Stephen Porges, , The Polyvagal Theory (2011). The neurological mechanism: how the autonomic nervous system - specifically the ventral vagal complex - creates the physiological conditions for social engagement, curiosity, and learning. Fear is not a metaphor. It is a measurable physiological state that forecloses the cognitive capacities we most need under pressure.

Guy Claxton, , Intelligence in the Flesh (2015). The most accessible account of embodied cognition - the argument that intelligence is not a property of the brain alone but of the whole body in relation to its environment. Claxton asks the question that IQ and EQ both avoid: what if intelligence is not primarily computational at all?

On the measurement error and its consequences

Stephen Jay Gould, , The Mismeasure of Man (1981, revised 1996). The definitive account of how IQ became the dominant measure of human worth, and the political and scientific misappropriations that made it so. Gould traces the Stanford-Binet’s role in American eugenics with forensic precision.

Daniel Goleman, , Emotional Intelligence (1995). Worth reading both for the genuine insight and for how quickly it was domesticated into a competency framework. The gap between what Goleman actually argued and what EQ became in corporate practice is itself a case study in the recapture of the relational by the reductive.

On organisational learning and human capability

Ruth Deakin Crick et al., , ‘Learning Power in the Workplace’, Journal of Management Development (2015). The peer-reviewed foundation of the Learning Power framework - structural equation modelling across sectors demonstrating that the eight CLARA dimensions predict adaptive capacity, performance, and sustained behaviour change.

Robert Kegan & Lisa Lachow Lahey, , Immunity to Change (2009). The most rigorous account of why organisations and individuals fail to change even when they genuinely want to. Kegan’s developmental framework maps directly onto the Learning Power dimensions and explains why most change programmes fail at the level of structure, not intention.

On AI and human capability

Anthropic, , Claude’s Constitution (2025). Anthropic’s published account of Constitutional AI - the attempt to encode ethical orientation into the architecture of the model rather than applying rules from outside. The philosophical alignment with WILD’s approach is not incidental.

Anthropic Economic Index, , ‘New Building Blocks for Understanding AI Use’ (2025). The empirical data on how AI is actually being used across industries and geographies - including the adoption gap and the distributional questions that Daniela Amodei raises in the Sixth Street podcast.

Jensen Huang has a gift for making the complex simple. The NVIDIA CEO’s description of AI as a five-layer cake, energy, chips, cloud infrastructure, AI models, and applications, has given executives, investors and policymakers a shared map of the territory.

It’s a good map, but to paraphrase Alfred Korzybski, maps leave things out, and what this one omits is the ingredient that determines whether the whole cake rises or not.

Not a sixth layer sitting on top. Something more like baking powder: invisible in the finished product, present through every layer, and entirely responsible for whether you end up with something edible or a flat, expensive, indigestible mess.

That ingredient is human infrastructure.

The cake, layer by layer

Huang’s framework starts at the bottom with energy: the power that makes everything else possible. Then chips: the specialised silicon that runs AI workloads. Above that, cloud infrastructure: the data centres and networks that host the compute. Then the models themselves: the large language models and multimodal systems that are the visible face of AI. And finally, applications: the products, services and workflows where economic value is actually created.

The logic is clean, with each layer dependent on the one below. The application layer is where Huang says the real benefit lands, where AI stops being a technology and starts being a capability.

But a capability for whom? Deployed by whom? Integrated, adapted, questioned, and directed by whom?

The answer is humans. And the question of whether those humans have the adaptive capacity, Learning Power and relational intelligence to work effectively alongside AI is simply not on the map.

Scoring the scorecard

Run the five layers as a quick geopolitical audit and the picture becomes uncomfortable, particularly for the UK.

The United States leads on models and capital. The race to Artificial General Intelligence is being run at a pace that even its own engineers admit outstrips any governance architecture. But Gillian Tett in The Financial Times has been persistent on a point that the AI headlines tend to obscure: American AI ambitions could be undone by the humdrum issue of power. Data centres demand extraordinary quantities of electricity. The grid is under strain. The energy layer, the foundation of Huang’s cake, is becoming a genuine constraint on US AI dominance. And the relentless focus on faster models means the human capability to use them wisely is an afterthought, not a design principle.

China is taking a different approach entirely. Practical over theoretical, open over closed, long-term over quarterly. Chinese AI investment is concentrated in health, education, agriculture and infrastructure, domains where human capability still matters, where the interface between person and system is the point, not an inconvenience. China is winning on the energy layer, particularly in green technology. And it has made a strategic choice about its human infrastructure that the West is only beginning to understand.

ByteDance (the company behind TikTok) built two versions of the same app. The version available inside China called Douyin has strict time limits for under-18s, no late-night access, and an algorithm that actively surfaces educational content. The version exported to the rest of the world was optimised for engagement and addiction. They knew exactly what the algorithm did and they chose not to do it to their own children.

The UK and Europe are, as Huang’s framework makes clear, laggards at almost every visible layer. Behind on energy transition. Dependent on US cloud providers. Marginal in frontier model development. On chips, the story is more complicated, and more revealing. Britain produced arguably the most important chip architecture in the world. ARM, headquartered in Cambridge, is the invisible foundation of virtually every smartphone on the planet and, increasingly, of AI data centres globally. Half of all compute shipped to the world’s top cloud providers in 2025 ran on ARM architecture. NVIDIA licenses it. Apple builds on it. Google, Amazon, Microsoft - all of them. And yet ARM is majority-owned by a Japanese conglomerate and listed in New York. This is a pattern - Britain created it and didn’t hold it. At the application layer, where economic benefit is supposed to land, the questions are whether there are businesses capable of building on these foundations, and workers capable of using them.

The bill has arrived

Research published by the Resolution Foundation, recently highlighted by John Burns-Murdoch in The Financial Times, found that Britain has the highest proportion of young people not in education, employment or training (NEETs) in the developed world. Nearly a million under-25s. Sixty percent have never held a paid job. A quarter cite health, including mental health, as the reason.

This is not simply a welfare statistic, it is a human infrastructure statistic. It describes the condition of the generation that will have to work alongside AI, the people at the application layer of Huang’s cake. And it tells us that layer is severely compromised.

The harvest of a decade of unregulated social media exposure to young people is now being counted in disengagement, mental health crisis and structural economic exclusion. Britain imported the addictive version of the algorithm and applied it to an entire generation. China protected its own.

The government’s response? The Home Office has launched a review of visas to attract top talent from abroad. Which is to say: the answer to a depleted domestic human infrastructure is to import someone else’s. A workaround, not a solution. It depletes another country’s sixth layer to compensate for the failure to invest in our own.

Starting my career at Procter & Gamble was instructive on this point. P&G has a longstanding principle that shapes everything from hiring to leadership development: build and promote from within. The logic is not sentimentality; it is that the capability you develop internally compounds over time in ways that imported capability never can. You can hire brilliant individuals but you cannot hire an organisational learning culture.

The grid that can’t be built

Nowhere is this more visible, or more consequential, than in Britain’s Great Grid Upgrade.

The £60 billion programme to overhaul the UK’s electricity transmission network is, adjusted for inflation, costing twenty times more than the construction of the original grid and supergrid combined. It is Critical National Infrastructure - the energy foundation that every other layer of Huang’s cake depends on. Without it, Britain’s AI ambitions are literally unpowered.

And it is struggling. Not primarily because of planning delays or pylon protests or the complexity of connecting Scottish wind to southern demand, though all of those are real. It is struggling because of a skills shortage that the energy union Prospect described in terms that bear repeating: “dire shortages,” “chronic understaffing,” “woefully under-resourced.” In their most recent survey, 82% of energy workers said staffing levels were too low. 69% reported tangible skills gaps in their organisation.

One specialist recruiter put the argument plainly: the biggest constraint on grid delivery is not generation capacity. It is “the growing shortage of specialist skills required to upgrade and operate the UK’s electricity grid. This is where delivery risk is quietly building.”

National Grid has recognised this. It launched what it calls a “pioneering enterprise model” - a Great Grid Partnership with supply chain partners - explicitly as a response to skills shortage. The physical infrastructure problem and the human infrastructure problem are, in their own framing, the same problem.

Technology infrastructure deploys in quarters. Human infrastructure develops over years. Planning both on the same timeline means you are structurally late on the variable that takes longest to mature. Britain is discovering this in real time, in the programme that most needs to succeed.

Not a layer, an ingredient.

So is human infrastructure the sixth layer of the cake, or something else?

The sixth layer framing is useful as an entry point. It says: you have left something out of your model, and that omission explains a great deal of what isn’t working.

But it’s deeper than that: human infrastructure isn’t additive, it runs through every layer. Energy systems need humans who can plan, build and adapt complex delivery programmes under conditions of radical uncertainty. Chip fabrication requires engineering cultures that sustain deep expertise across generations. Cloud infrastructure depends on organisations capable of operating and evolving systems that were never fully designed. Models need humans who can interrogate outputs rather than simply accept them. And applications, where the value is supposed to land, are only as good as the human capability to integrate AI into real work, real decisions, real relationships.

For the UK this is not a new vulnerability. We discovered North Sea oil in the same decade Norway did. Around twenty years later, in 1990, Norway built the Government Pension Fund. Now worth over $1.7 trillion, Norway developed the institutional capability to think in generations rather than quarters whilst Britain extracted and spent. The human infrastructure to make a different choice - the policy capability, the long-term governance instinct, the wisdom to resist short-term extraction - was the difference. ARM tells the same story in a different decade. The capability to create is not the same as the capability to retain.

Before the crisis, or after

When I published an earlier piece arguing that AI is the wrong unit of analysis, it drew hundreds of responses. One comment stood out: Sener Cem Irmak, founder of Koru Impact, an evidence-first adaptive systems consultancy working with C-suites and private equity, offered a diagnosis of the governance problem at the heart of this argument.

“We’ve seen this pattern before,” he wrote. “Risk management spent decades in this liminal state: too technical for the board, too strategic for operations, too cross-cutting for any function. Everybody’s concern, nobody’s discipline. Then Basel II, COSO, and 2008 forced it into its own governance architecture: dedicated reporting lines, board-level accountability, measurable KRIs.

Human infrastructure is at that inflection point, but the question isn’t which function should own it. The question is whether organisations build a dedicated governance architecture for adaptive capacity before a crisis forces it, or after.”

The risk management precedent is illuminating. Before 2008, risk was structurally orphaned in most organisations: acknowledged in principle, unowned in practice, measured inconsistently if at all. The financial crisis didn’t create the risk management discipline, but it forced it into governance architecture that boards could no longer ignore. Chief Risk Officers acquired real authority. Measurable frameworks became mandatory. The function that had been everyone’s concern and nobody’s discipline became a defined professional field with its own reporting lines, its own metrics, its own seat at the table.

Human infrastructure is in exactly that pre-2008 position today. Acknowledged everywhere, owned nowhere. Every organisation depends on the adaptive capacity of its people. Almost none measures it, governs it, or funds it with the seriousness the dependency deserves.

Sener added a sequencing point that belongs at the centre of every infrastructure investment conversation: technology deploys in quarters, human infrastructure develops over years. Planning both on the same timeline means you are structurally late on the variable that takes longest to mature. The Great Grid Upgrade is discovering this. Britain’s AI ambitions are about to.

The measurability work already exists. For nearly three decades Professor Ruth Crick has been leading research into “Learning Power”: not skills, not qualifications, but the adaptive capacity that allows individuals and organisations to keep learning as conditions change. The disposition to engage with complexity rather than retreat from it. The relational intelligence to work across difference. The capacity to hold uncertainty without collapsing into either false certainty or paralysis.

Professor Crick’s Learning Power framework has been doing precisely what Sener identified as the precondition for governance: making the unmeasured measurable, making the unfunded fundable. You cannot govern what you cannot measure. And you cannot fund what you cannot govern.

The only question, as Sener framed it, is whether organisations and nations build the governance architecture before the crisis, or after. History suggests most will wait. But those that don’t will have a structural advantage that compounds, quietly, over years, in exactly the way that human infrastructure does.

What the AI billionaires can’t see

There is another relevant thread in recent commentary worth naming. The men who built the current wave of AI, the founders, the funders, the stage-fillers, are exhibiting signs of a particular kind of anxiety. Not about Artificial General Intelligence in the abstract. About something more personal: the cognitive style that made them successful is now being replicated by their own product. Linear pattern recognition. The ability to process information faster than others, spot trends earlier, optimise systems. The intelligence that gets you into Stanford, through Y Combinator, onto a stage in front of a quarter million people. That intelligence is now a commodity. They know it. They can feel it. And they’re projecting their anxiety on to us.

Jensen Huang himself put it plainly when asked who was the most intelligent person he’d ever met. He declined to name anyone, because the question itself revealed the problem. “The definition of smart is someone who solves problems, technical,” he said. “But I find that’s a commodity, and we’re about to prove that Artificial Intelligence is able to handle that part easiest.” Software programming, he noted, was supposed to be the ultimate marker of intelligence and it was the first thing AI solved. So the definition of smart, he concluded, needs to change entirely.

And yet, when the people building AI are asked not what the technology can do, but what their own children will need, something shifts. The Wall Street Journal recently asked several leading AI figures exactly that question. Ethan Mollick pointed toward flexibility and broad education: the things you’d invest in “in any time of uncertainty.” Daniela Amodei, President of Anthropic, was more direct: “When I think about what my kids will need as they get older, it’s human qualities: the ability to relate, to empathise and be around other humans. What’s not going to be replaceable is how you treat other people, how well you communicate with them, how kind you are.”

They are describing, for their children, exactly what they are failing to build for their organisations and their nations. They can see the missing ingredient, they just haven’t yet worked out how to bake with it.

What they cannot feel, because they have never had to develop it, because it was never what got them on the stage, is the capability that AI cannot replicate: genuine curiosity, relational depth, embodied judgement, the capacity to learn in conditions of ambiguity, the ability to ask the question that doesn’t yet have an answer.

That is what human infrastructure names. Not soft skills. Not culture. The human substrate - the measurable, developable adaptive capacity of people and organisations to engage with what they don’t yet know.

Jensen Huang is right about the layers. He’s right that the application layer is where value lands. He’s right that energy is the foundation.

But the nation most likely to win the AI era is not the one with the best chips or the biggest models. It is the one that has invested, seriously, measurably, at scale, in the human capability to use all of the above wisely.

Britain is not currently that nation, but it could be.

The teaspoon of baking powder in every cake never gets the credit, but take it out and see what you’re left with.

—

A note on sources

Jensen Huang’s five-layer framework was articulated at Davos and expanded in conversation with BlackRock CEO Larry Fink (blog on NVIDIA’s website here). His remarks on the redefinition of intelligence appeared on the A Bit Personal Podcast with Jodi Shelton.

Alfred Korzybski was a Polish-American scientist and philosopher working in the 1930s who coined the phrase “the map is not the territory”. He reminds us that our mental models of the world are not the same as the world itself - maps help us simplify complexity. Excellent read on mental models by Shane Parrish here.

The Wall Street Journal’s interviews with Ethan Mollick and Daniela Amodei on AI and the future of education were shared widely, including by Dr Will Van Reyk Deputy Headmaster of St Paul’s School.

Link to The Resolution Foundation’s NEET research here - thanks to John Burns-Murdoch for reporting on it in The Financial Times.

The Great Grid Upgrade skills shortage data comes from Prospect union workforce surveys and specialist energy sector commentary.

The Douyin/TikTok comparison is documented across multiple independent investigations into ByteDance’s domestic and international product strategies.

The observation that the AI billionaires’ anxiety stems from the obsolescence of linear pattern recognition was made with characteristic precision by Abi Awomosu on Substack.

The Learning Power framework is the result of twenty-six years of research led by Professor Ruth Crick (originally at the University of Bristol) founder of WILD Learning. Crick’s research has been validated across more than 100,000 individuals and 190 organisations using Structural Equation Modelling (the rigorous statistical method) to confirm the robustness of its 8 dimensions measurement framework. WILD stands for Work Integrated Learning Design and is the framework that scaffolds “learning journeys”.

AI will amplify whatever system it’s placed into.

If your organisation is optimised for short-term extraction - faster decisions, leaner headcount, tighter control - AI will accelerate that. More dashboards. More confident conclusions at speed. More of what you already had, delivered faster.

If your organisation is optimised for learning, for interpreting complexity together, for coordinating across boundaries, for adapting faster than your competitors - AI will amplify that instead.

The question every leader needs to answer isn’t “how do we adopt AI?” It’s simpler and more uncomfortable than that: what have we actually built - and do we want more of it?

Most AI strategies never ask this question, because they get stuck much earlier in the conversation.

The question nobody’s asking

The best framework I know for cutting through strategic noise comes from AG Lafley and Roger Martin’s Playing to Win, the strategy model developed at P&G in the early 2000s and now used across FTSE and Fortune 500 boardrooms. It asks five questions, in sequence:

1. What is our winning aspiration?

2. Where will we play?

3. How will we win?

4. What capabilities must be in place?

5. What management systems are required?

Most AI conversations are stuck on questions one to three. Should we adopt AI? Where should we deploy it? What’s the business case? These are legitimate questions, but they are not where strategy becomes real. Strategy only becomes real at question four.

Roger Martin himself, writing in 2023, warned against this: “Don’t spend disproportionate time on first three boxes. That is a classic strategy mistake.” He describes questions four and five as the essential reality check, the test of whether your strategy is actually executable. Without them, he argues, you have at best a draft.

But I’d go further. In an AI-enabled world, questions four and five are not just a reality check, they are where the strategic work actually lives. The capability to learn, coordinate and adapt at speed is not the validation of your strategy. It is the strategy.

What capabilities must be in place? This is the question that separates organisations that will genuinely benefit from AI from those that will spend a lot of money going faster in the wrong direction. And almost nobody is asking it.

Here’s why it matters so much right now: AI changes the answer to question four fundamentally. The capability that used to matter most - analytical horsepower, process discipline, information advantage - is increasingly commoditised. Daniel Pink saw this coming twenty years ago, arguing in A Whole New Mind that the future belonged not to left-brain analytical thinking but to the right-brain capacities of synthesis, empathy, and meaning-making. AI has simply accelerated the timeline. What AI cannot do is coordinate, interpret, and learn together at speed. That remains irreducibly human.

Which means the capability gap that will determine who wins in an AI-enabled world is not technical, it’s human. And most organisations are investing as though the opposite were true.

Question five is where hope enters

What management systems are required? In most organisations, this means dashboards, governance frameworks, reporting lines; the apparatus of control. And for AI investment, it increasingly means sophisticated technology metrics: adoption rates, productivity gains, cost per output.

None of that is wrong, but it is incomplete, and the incompleteness is dangerous.

If the capability that matters most is human adaptive capacity, then your management systems need to be able to see it, track it, and develop it. And here is where most leaders hit a wall. Human capability feels unmeasurable because it’s distributed, emergent, relational. Capital markets prefer linear cause and effect, so the investment case collapses before it’s made.

This is exactly the problem that Professor Ruth Crick’s research solves. Her work identifies what she calls Learning Power, eight dispositional capacities that determine whether individuals and organisations can actually engage with uncertainty, novelty, and genuine challenge:

Mindful Agency. Sense Making. Curiosity. Creativity. Hope & Optimism. Belonging. Collaboration. Orientation to Learning.

These are not soft skills. They are not a competency framework. They are measurable, developable capacities that describe how humans actually come to know, decide, and act when embedded in complex living systems. And they are almost entirely absent from every AI leadership framework currently on the market.

Question five, properly answered, looks like this: pair every AI capital investment with a Learning Power baseline. Measure it. Develop it. Track it alongside your technology metrics. Then you have a management system that can actually see the whole system, not just the machine layer.

That is not a talent programme. That is strategy.

So what does this mean on in the real world?

The leaders who will navigate this well are not necessarily the ones with the biggest AI budgets. They are the ones who ask a different set of questions.

Not just what is our AI strategy? But, what does our organisation need to be capable of for that strategy to work?

Not just what is the ROI of this AI investment? But, what is the organisational risk if we don’t upgrade our human system at the same time?

In practice, this means three things.

First, elevate your Chief People Officer to genuine strategic partner on AI transformation, not downstream of the technology decisions, but at the table where they are made. The human capability question is not an HR question. It is a strategy question.

Second, pair every significant AI capital investment with a human capability baseline. Measure where your organisation sits on the dimensions that actually determine whether AI generates collective intelligence or collective confusion. You cannot manage what you cannot see.

Third, redesign what you reward. Most incentive structures still reward individual output and short-term delivery. If the capability that matters is cross-boundary learning, coordination quality, and adaptive speed, then your incentives are pointing in the wrong direction. That is not a culture problem. It is a systems design problem, and it has a systems design solution.

None of this is complicated, but some of it is uncomfortable, particularly for leaders who have built their identity around being the person with the answers. The organisations that will genuinely benefit from AI are not the ones with the most confident leaders. They are the ones with the most capable learners.

The risk is real. So is the answer.

We are spending billions on AI capability and almost nothing on the human capability to use it wisely. That is not a talent gap, it is a systems risk and it is hiding in plain sight on every board agenda that discusses AI adoption without once asking question four.

The good news is that this is not an intractable problem. It is not a values problem or a culture problem or a leadership personality problem. It is a measurement problem, and measurement problems have solutions.

We know what human adaptive capacity looks like. We know how to assess it, develop it, and track it over time. We know which dispositions determine whether an organisation learns faster than its competitors or just generates faster mistakes with greater conviction. The frameworks exist. The evidence base exists. The tools exist.

What has been missing is the strategic will to treat human infrastructure with the same rigour we apply to technology infrastructure.

AI will amplify whatever system it’s placed into. The organisations that thrive will be the ones whose leaders had the courage and the clarity to ask not just what are we building with AI? But, what are we building for AI to work within?

That question is available to every leader reading this. Right now. Before the next investment decision. Before the next board presentation. Before the next transformation programme is signed off.

The system you build is a choice. And it is still yours to make.

********

References & Further Reading

The following thinkers and works have shaped the argument in this piece, each is worth going deeper on.

AG Lafley & Roger Martin: Playing to Win: How Strategy Really Works (2013) The source of the five-question strategic cascade used throughout this piece. Developed at Procter & Gamble in the 2000s and now one of the most widely used strategy frameworks in global business. Its power lies in the sequencing and the iterative nature of asking the questions, and in making question four, capability, a strategic rather than operational question. Required reading for anyone serious about strategy.

Daniel Pink: A Whole New Mind: Why Right-Brainers Will Rule the Future (2005) Written twenty years before the current AI moment, Pink argued that the analytical, left-brain capabilities most prized by organisations were becoming commoditised and that synthesis, empathy, creativity, and meaning-making would become the decisive human advantages. AI has simply compressed his timeline. Remarkably prescient.

Professor Ruth Crick - Learning Power framework Ruth Crick was Professor of Learning Analytics and Educational Leadership at the University of Technology Sydney, an academic at Bristol University and founder of WILD Learning. Her research identifies eight measurable, developable dispositional capacities - Learning Power - that determine whether individuals and organisations can engage productively with complexity, uncertainty, and genuine challenge. This is the empirical foundation for treating human adaptive capacity as a strategic and measurable asset. More at wildlearning.com

Arie de Geus: The Living Company (1997) De Geus studied companies that had survived for 500 years or more and concluded that the only sustainable competitive advantage is the ability to learn faster than your competitors. Written before the internet, let alone AI and more relevant now than ever. De Geus used flocking - the emergent collective movement of birds - as a metaphor for how living companies sense and adapt together. No central controller. Intelligence distributed across the whole. The murmuration image at the head of this piece is a nod to that idea.

Jody Hoffer Gittell: The Southwest Airlines Way (2003) Gittell’s research on relational coordination - the combination of shared goals, shared knowledge, mutual respect, and frequent high-quality communication - showed that Southwest’s competitive advantage wasn’t aircraft or routes. It was how people worked together across boundaries. Her framework explains why learning speed in complex systems is fundamentally a relational question.

However, there’s a poignant postscript: Southwest Airlines, the very company Gittell studied, has since become an illustration of what happens when relational coordination is sacrificed for short-term extraction. Following the death of founder Herb Kelleher and pressure from activist investor Elliott Management, Southwest abandoned the open seating model and cultural practices that defined its advantage in favour of cost reduction and revenue maximisation. The Christmas 2022 IT collapse - rooted in years of underinvestment - was not a technology failure, it was a systems failure. Gittell’s thesis, if anything, is proved more powerfully in the breach. A sad example of what happens when you amplify for short term returns.

He wasn’t warning about the technology. He was warning about us.

I can’t stop thinking about that line, not because it’s alarming, though it is, but because it’s asking a human question, not a technical one. Maturity isn’t something you acquire by upgrading your software. It isn’t a capability you can buy, deploy or prompt-engineer your way into. It’s the accumulated result of learning to hold uncertainty, to stay in relationship under pressure, to be changed by experience rather than merely informed by it.

And yet the entire conversation about AI leadership is structured around the wrong question.

Boardrooms, business schools, executive coaching programmes (including some very good ones) are asking “how do we become AI-ready?” They’re hiring Chief AI Officers, running digital literacy programmes, sending leaders on prompt engineering workshops. They’re treating an adaptive challenge as a technical problem, which Ron Heifetz identified decades ago as the central failure of leadership: reaching for a known answer in the system because the alternative, sitting with genuine uncertainty, is simply too threatening.

I’ve spent thirty years watching this pattern repeat. SAAP implementations. ERP systems. Agile transformation. Digital disruption. Every wave of powerful new technology exposes the same underlying deficit, and every time we respond by focusing on the tool rather than the human infrastructure required to use it wisely. The tool changes. The gap persists.

AI doesn’t create that problem - it reveals it, accelerates it, and, if Amodei is right about the trajectory, makes it impossible to keep ignoring.

* * *

The left-hemisphere trap

Iain McGilchrist’s work on the divided brain offers something precise here that most leadership conversations miss entirely. The left hemisphere of the brain - brilliant, systematic, certain - doesn’t just prefer clarity. It actively suppresses what it can’t categorise. It mistakes the map for the territory. It takes a simplified model of reality and treats it as reality itself.

Our dominant organisational model is essentially a left-hemisphere machine. Hierarchies, KPIs, governance frameworks, planning cycles, reporting structures - all of it is exquisitely designed for categorisation, control and prediction. It excels at what Dave Snowden’s Cynefin calls complicated problems: the kind where there is a known answer somewhere in the system, where expertise applies, where cause and effect are visible.

But the problems AI is being deployed into - infrastructure delivery, healthcare transformation, strategic change at scale - are complex, not complicated. Cause and effect are only visible in retrospect. No individual however credentialled, can lead their way through that from the top.

So here is what happens when you introduce AI into a left-hemisphere organisation: the organisation grabs it and uses it to generate more certainty, faster. More data. More dashboards. More confident decisions at speed.

Which in a genuinely complex system doesn’t produce better outcomes. It produces faster mistakes made with greater conviction.

* * *

The immune system always wins

There’s a deeper reason this pattern is so hard to break, and it has nothing to do with intelligence or intention.

Bob Kegan and Lisa Lahey’s research on immunity to change shows that even when people receive life-threatening medical diagnoses and are told exactly what behavioural changes would save them, fewer than 20% make those changes. Not because they don’t understand. Not because they don’t care. But because they have competing commitments - deeply held, often unconscious assumptions about what keeping themselves safe requires - that are more powerful than any stated intention to change. The immune system rejects the new heart even when the patient needs it to survive.

In organisational cultures this dynamic runs even deeper. Fiona Robertson captures it exactly: “We behave our way to belonging.” In a strong Clan culture - and most large organisations are stronger Clan cultures than they would care to admit - the rules of belonging are more powerful than any change programme. Leaders aren’t resisting AI transformation irrationally. They’re protecting something real: their status, their certainty, their sense of being the person with the answers.

David Rock’s SCARF model shows precisely why AI-era leadership triggers this response so acutely. Status is threatened when you don’t have the answers. Certainty collapses in genuinely complex environments. Autonomy is undermined by algorithmic systems you don’t fully understand. Relatedness frays in distributed, hybrid working. Fairness feels violated when AI-driven decisions are opaque. All five threat responses activate simultaneously. The workshop happens. The credential is acquired. And the culture quietly protects itself back to equilibrium.

This is why the AI leadership gap isn’t a training problem. It’s an immunity to change problem. And until we treat it as such - surfacing competing commitments, working with the belonging dynamic, regulating adaptive pressure rather than eliminating it - we will keep producing leaders who are credentialled but not capable.

* * *

The gap we’re not measuring

What organisations actually need to navigate AI-era complexity has a name. Professor Ruth Crick’s decades of research identifies what she calls Learning Power - eight dispositional capacities that determine whether individuals and organisations can engage productively with uncertainty, novelty and genuine challenge:

Mindful Agency: acting purposefully while remaining adaptive. Sense Making: constructing shared meaning from ambiguous, conflicting information across boundaries. Curiosity: staying in inquiry rather than rushing to closure. Creativity: generating new possibilities rather than choosing between existing options. Hope and Optimism: the resilience to keep engaging when outcomes are uncertain and timescales are long. Belonging: a felt sense of being part of something larger than your employer. Collaboration: working across difference, co-generating rather than merely coordinating. Orientation to Learning: the willingness to be changed by the process itself.

These are measurable. They are developable. And they are almost entirely absent from every AI leadership framework currently on the market.

Instead, those frameworks are built around individual skill acquisition - a complicated-domain response to a complex-domain challenge. The mismatch between the capability being built and the environment it’s being built for is where organisations will fail. And they won’t even know why, because they’ll be measuring the wrong things.

We are spending billions on AI capability and almost nothing on the human capability to use it wisely. That is not a talent gap. It is a systems risk - and it is entirely measurable if you know what to look for.

* * *

Maturity, revisited

Matt Shumer, an AI founder writing recently about the pace of change, described the current moment as being in the “this seems overblown” phase of something much bigger than most people are prepared for. He’s right about the acceleration. He’s right about the urgency. His advice - use the tools seriously, adapt fast, build the habit of learning - is sound as far as it goes.

But his entire framework for what to do is still structured around the individual and the technical. Learn more. Prompt better. Get ahead of the curve. It is, in Heifetz’s terms, a technical response to an adaptive challenge - and a familiar one.

What neither Shumer nor most of the AI leadership conversation have quite got to is this: the question was never whether the technology works. It’s whether the human systems around it are capable of using it wisely. That question doesn’t get answered by better prompts or faster adoption. It gets answered by the slow, difficult, unsexy work of building genuine human infrastructure - the Learning Power, the relational capacity, the adaptive leadership capability that allows organisations to navigate complexity rather than just accelerate through it.

Amodei calls this a test of maturity. He’s right. But maturity, in this sense, is not a disposition. It is a capability. It is built, deliberately, over time, through exactly the kinds of experiences and conditions that our left-hemisphere organisations are currently designed to prevent.

The AI conversation has been asking the wrong question. The right question is not “are we AI-ready?”

It is: do we have the human infrastructure to be ready for anything?

And then, honestly: do we even know how to measure that?

* * *

Victoria Ferrier is a strategic advisor working at the intersection of human capability, complex systems and organisational performance. She is collaborating with Professor Ruth Crick on the application of Learning Power to large-scale infrastructure delivery. This piece was developed in active collaboration with Claude - Anthropic’s AI assistant - as a demonstration of the kind of human-AI partnership the argument advocates for.

References & Sources

Amodei, D. (2025). Machines of Loving Grace. Anthropic. [The essay underpinning the maturity warning that opens this piece]

Shumer, M. (2026). Something Big Is Happening. shumer.dev. [The piece that prompted this response - right about the urgency, asking the wrong question about what to do]

McGilchrist, I. (2009). The Master and His Emissary: The Divided Brain and the Making of the Western World. Yale University Press.

Heifetz, R. & Linsky, M. (2002). Leadership on the Line: Staying Alive through the Dangers of Change. Harvard Business Press.

Kegan, R. & Lahey, L. (2009). Immunity to Change. Harvard Business Press.

Robertson, F. (2016). The Rules of Belonging. [Source of “we behave our way to belonging”]

Rock, D. (2008). SCARF: A Brain-Based Model for Collaborating With and Influencing Others. NeuroLeadership Journal.

Snowden, D. & Boone, M. (2007). A Leader’s Framework for Decision Making. Harvard Business Review. [The Cynefin Framework]

Crick, R.D. et al. (2004). Characterising Learning Power. University of Bristol. [The foundational research behind the eight Learning Power dimensions and the foundation of Professor Crick’s WILD framework for building human infrastructure]

The question keeps surfacing in different forms: Is it inevitable? Has the ship already sailed? Are we merely passengers now, watching a technological future arrive whether we consent to it or not?

Brand strategist Zoe Scaman recently argued that this sense of inevitability is not a neutral assessment of reality, but a story: one that functions to shut down dissent and foreclose imagination. When people speak of AI as inevitable, she suggests, what they are really saying is: stop asking questions; get in line. It is delivered with the tone of prophecy, as though technology obeys the laws of gravity rather than the choices of humans with power.

I share her refusal of that framing, and I want to push the question deeper still. What if AI becomes “inevitable” not because it must, but because we lack the inner and collective capacity to choose otherwise?

From this perspective, the AI moment is not primarily a technological crisis. It is a developmental one. And more precisely, it is a crisis of human Will.

Inevitability as a Failure of Human Development

Across several decades of research in adult development, a striking and uncomfortable finding recurs: only a small minority of adults - often estimated at under 10% - have developed the cognitive and emotional capacity required to lead effectively in conditions of high complexity, ambiguity, and rapid change.

This research distinguishes between horizontal development (the acquisition of new skills, knowledge, and competencies) and vertical development, which refers to how adults make meaning of the world itself. Vertical development reflects the capacity to hold multiple perspectives, tolerate paradox, think systemically, and remain ethically grounded under pressure.

The challenge we face is not a shortage of intelligence or expertise. It is a mismatch between the complexity of the systems we have built and the developmental maturity of the humans charged with stewarding them.

Artificial intelligence dramatically intensifies this mismatch. AI systems amplify intention at scale, but they do not generate intention. AI accelerates whatever values, incentives, and assumptions already dominate the systems in which they are embedded. When those systems are governed primarily by short-term profit, competitive dominance, and control, AI does not challenge those logics. It perfects them.

If AI feels inevitable, it may be because we are asking a developmentally underprepared population to make wise choices under unprecedented conditions of complexity.

A Crisis of Will

Roberto Assagioli, the founder of psychosynthesis, defined the Will not as brute force or domination, but as the capacity for conscious choice in service of a purpose. A volitional act, he argued, requires a clear vision of an aim and the inner coherence to pursue it.

By this definition, our moment is marked by an extraordinary imbalance: vast technological power paired with a fragile and underdeveloped collective Will.

We are highly capable of optimisation, but far less capable of discernment. We can build systems that predict behaviour at scale yet struggle to ask toward what ends those systems should serve. In the absence of a cultivated Will, we default to the prevailing values of efficiency, growth, extraction, and control.

This is not a moral failing so much as a developmental one. Without practices that strengthen the Will - discernment, self-regulation, and ethical sense-making - power drifts toward the loudest incentives and the most concentrated interests.

The Left Hemisphere World

The philosopher and psychiatrist Iain McGilchrist offers a useful lens for understanding this drift. His work on hemispheric imbalance suggests that Western societies have become increasingly dominated by left-hemisphere modes of attention - those that privilege abstraction, certainty, measurement, and control - at the expense of right-hemisphere capacities for context, relationship, embodiment, and meaning.

AI is, in many respects, a left-hemisphere technology. It excels at decomposition, categorisation, optimisation, and prediction. These are powerful capabilities. But without right-hemisphere leadership, without attention to lived context, ethical consequence, and human relationship, such tools tend toward reductionism and coercion.

A civilisation already biased toward abstraction and control will build machines in its own image.

Authoritarianism as Coherence Under Threat

This brings us to the question of authoritarianism and power.

One of the most important insights in Scaman’s essay is her recognition that inevitability narratives serve power. They naturalise concentration of control, particularly when AI infrastructure is owned by a small number of billionaires whose interests are not aligned with democratic participation or the common good.

But authoritarianism does not succeed solely because of elite manipulation. It succeeds because it offers coherence in moments of fear and disorientation.

In the US, the MAGA story did not win because it was factually accurate. It won because it metabolised fear into purpose. It named a threat - wrongly and destructively - but in doing so, it offered orientation. A scapegoat, a banner, and a rhythm to march to.

“There is a threat,” the story says. “And here is who you are, and here is where you belong.”

The UK has its own version of this pattern. Writing in the Financial Times, Chris Giles recently observed that “the UK economy is not nearly as bad as you’ve been told”, that Bank of England forecasts predicting the longest-ever recession in 2022 were wrong by a significant margin, and that a persistent pessimistic bias in national accounts has fed a doom narrative that the data doesn’t support. We are hardwired for negativity bias and authoritarian stories know how to exploit that wiring.

By contrast, the progressive response to systemic instability has often been technocratic. Fact-checking. Policy briefs. Programmatic solutions. An aesthetics of correctness. But when systems are shaking, people do not primarily want more information. They want location.

The Need for an Alternative Story

Sociologist Dr. Tressie McMillan Cottom has been particularly influential in shaping my thinking here. In a recent podcast conversation, she speaks not in the language of grand ideology, but of small acts, local solidarities, and community-level responses to power. She resists abstract utopianism and instead emphasises lived, relational alternatives: people making meaning and exercising agency where they are.

What she’s describing is communitarianism.

Communitarianism offers a counter-story to authoritarianism, not one rooted in exclusion or domination, but in mutual obligation, shared responsibility, and the common good. It answers a different question than liberal individualism. Not merely what are my rights? but who holds me and whom do I hold?

Authoritarianism offers belonging through exclusion. Communitarianism offers belonging through participation.

This distinction matters enormously in an AI-shaped future. Technologies that centralise surveillance and prediction thrive in cultures of fear and fragmentation. Technologies governed by communities, norms, and shared accountability require a different psychological and moral substrate.

The Future Is Not Written

If authoritarian AI futures feel inevitable, it may be because they are telling a story that meets people where they are, while alternative visions are still being communicated as if people are calm, resourced, and developmentally unthreatened.

The answer is not better facts alone. It is a better story: one rooted in communitarian belonging, developmental maturity, and a reclaimed human Will.

This is not abstract work. Professor Ruth Crick’s decades-long research into resilience suggests that it requires building specific human capacities rooted in agency. In the context of “inevitability” people need to learn to recognise authoritarian patterns when they appear, to interrupt them without collapsing into either compliance or reckless confrontation, to replace them with pro-social alternatives, and to embed those alternatives in systems rather than relying on heroic individuals.

AI does not decide the future. Billionaires do not decide the future. Stories, capacities, and choices do.

The future will be shaped by whether we invest not only in technologies, but in humans capable of discernment, relationship, and collective responsibility.

Inevitability is a claim. And claims can be refused.

In future writing, I’ll explore where this developmental work is already happening, and what it means to steward AI through relationship rather than alignment.

References

Scaman, Z. (2025). Is It Inevitable? Musings of a Wandering Mind.

Assagioli, R. (1973). The Act of Will. Viking Press.

Centre for Creative Leadership. (2025). Vertical vs. Horizontal Development: Why Your Leaders Need Both to Succeed.

Kegan, R., & Lahey, L. L. (2009). Immunity to Change. Harvard Business Press.

Torbert, W. R. (2014). Action Inquiry. Berrett-Koehler.

McGilchrist, I. (2009). The Master and His Emissary. Yale University Press.

Giles, C. (December 2025). The UK economy is not nearly as bad as you’ve been told. FT Weekend.

McMillan Cottom, T. (2025). We Can Do Hard Things [Podcast appearance].

Fabian Society. (2023). For the Common Good: A New Communitarianism.

Professor Ruth Crick. (2015). Developing Resilient Agency in Learning: The Internal Structure of Learning Power

As CFOs and Finance Directors, we’re familiar with the promise - and pressure - of AI transformation. We’ve approved the budgets, reviewed the business cases, and tracked the implementation timelines. Yet many of us are witnessing a troubling pattern: the technology delivers, but the transformation doesn’t.

The issue isn’t the AI itself. It’s the human infrastructure surrounding it. And this presents both a strategic opportunity and a financial imperative for finance leaders.

Why AI Adoption Is a CFO Issue

Recent research by Professor Ruth Crick and colleagues reveals that AI adoption fails when treated as a technology project rather than a learning challenge. For finance leaders, this insight reframes our role entirely.

We’re not just funding technology; we’re investing in our organisation’s capacity to learn, adapt, and evolve. The question isn’t simply “What’s the ROI on AI tools?” but “Have we built the human capability to realise that return?”

The answer increasingly determines whether our AI investments generate value or languish as underutilised assets on the balance sheet.

The Hidden Cost of Having Low Learning Capacity

Consider the true cost of failed AI adoption:

• Sunk technology investments with minimal productivity gains

• Employee disengagement as initiatives stall and enthusiasm wanes

• Competitive disadvantage as more adaptive competitors pull ahead

• Opportunity costs from delayed transformation

These costs rarely appear in project budgets, yet they dwarf the initial investment. They stem from treating AI as a capital expenditure when it’s actually an organisational capability issue.

The organisations succeeding with AI have recognised something fundamental: they’re not just implementing tools; they’re building what’s called Learning Power—the collective capacity to learn, unlearn, and relearn at scale.

What Is Learning Power and Why Should CFOs Care?

Learning Power comprises eight measurable dimensions that determine how individuals and teams engage with complexity and change:

• Mindful Agency – self-leadership and ownership of learning

• Curiosity and Creativity – drive to explore and innovate

• Sensemaking – ability to interpret complexity and discern meaning

• Hope and Optimism – emotional resilience and confidence

• Belonging and Collaboration – relational trust for collective learning

• Orientation to Learning – openness to feedback and the unknown

When these dimensions are strong, AI adoption accelerates. When they’re weak, even the best technology struggles.

Research by WILD Learning found that teams with higher learning power scores adopted AI-enabled tools 40% faster and sustained usage far longer. They weren’t more technical—they were more adaptive.

For CFOs, this represents a quantifiable leading indicator of transformation success.

A Case for Investment: Hunter Water

Hunter Water faced a critical challenge: consumer demand would exceed supply within 20 years. Rather than treating this as purely an infrastructure problem, they invested in building learning power across the organisation.

The results demonstrate the financial impact of this approach:

• Water leakage fell by 34% over three years

• Community usage dropped by 15% more than expected in year three

• Staff engagement rose significantly

• A culture of innovation became embedded across the company

This wasn’t achieved through technology alone, but through developing the human infrastructure - agency, curiosity, sensemaking, and collaborative learning - that enabled people to change how they worked and thought.

The CFO’s Strategic Levers for Building Learning Power

As finance leaders, we have unique influence over the conditions that cultivate learning power. Here are five practical ways to deploy that influence:

1. Reframe Capital Allocation

Move beyond viewing AI as purely a technology investment. Include “learning infrastructure” as a distinct budget category alongside digital infrastructure.

This means allocating resources to:

• Learning power diagnostics and analytics

• Leadership development in facilitating learning

• Peer-coaching and collaborative learning programmes

• Time and space for reflection and experimentation

When evaluating AI business cases, explicitly assess and cost the learning capacity required to realise projected benefits.

2. Redesign Performance Metrics

Traditional AI adoption metrics - system usage rates, tickets processed, or productivity outputs - are necessary but insufficient. They measure activity, not capability.

Champion the inclusion of learning indicators in transformation scorecards:

• Learning power assessment scores across teams

• Psychological safety indices

• Innovation experiment velocity

• Cross-functional collaboration metrics

• Rate of integration of AI into daily work practices

These leading indicators predict whether transformation will stick far better than lagging output metrics alone.

3. Model Adaptive Leadership

Finance teams often set the cultural tone for the organisation. Our response to AI adoption signals what’s valued and acceptable.

Demonstrate learning-powered leadership by:

• Publicly experimenting with AI tools in finance processes

• Sharing what you’re learning - including failures and uncertainties

• Creating reflective practices in FLT and board discussions

• Asking “What did we learn?” alongside “What did we deliver?”

• Allocating time for your team to explore, experiment, and make sense of AI

When finance leads with curiosity and adaptation, permission cascades through the organisation.

4. Invest in Relational Intelligence

AI adoption depends on trust, safety, and belonging - what researchers call “relational intelligence.” These aren’t soft factors; they’re the operating system for collective learning.

As CFOs, we can:

• Sponsor cross-functional learning communities around AI adoption

• Fund facilitated reflection and dialogue spaces

• Build peer-coaching capability across the organisation

• Measure and reward collaborative innovation, not just individual output

• Ensure AI implementation teams include learning specialists, not just technologists

The ROI on relational infrastructure often exceeds that of technical infrastructure.

5. Create Learning-Aligned Governance

Traditional programme governance focuses on milestones, deliverables, and risk management. Learning-powered governance adds a parallel track.

Establish governance that asks:

• How is learning power evolving across affected teams?

• Where are curiosity, agency, or sensemaking declining?

• What’s emerging that we hadn’t anticipated?

• How are we adapting our approach based on what we’re learning?

This creates an evidence-based feedback loop for transformation, allowing course correction before problems become crises.

From Cost Centre to Strategic Asset

Here’s the paradigm shift for finance leaders: learning power isn’t an operational expense; it’s strategic infrastructure.

Just as we wouldn’t deploy new ERP systems without adequate data architecture, we shouldn’t deploy AI without adequate learning architecture.

The organisations that will win in the AI era aren’t those with the most sophisticated algorithms - they’re those with the most adaptive people. And adaptiveness at scale requires investment in learning cultures.

Measuring What Matters

WILD Learning’s analytics platform allows organisations to measure and track learning power profiles over time, providing real-time visibility into the human side of transformation.

This creates something most AI programmes lack: actionable data on the readiness and capability of your people to integrate change.

For CFOs, this offers three advantages:

• Earlier risk identification when teams show low learning readiness

• Better resource allocation to areas requiring capability development

• Quantifiable evidence of transformation progress beyond technology deployment

When learning power data is linked to business outcomes - as at Hunter Water - it builds the business case for continued investment in human infrastructure.

The Bottom Line for Finance Leaders

AI will amplify whatever systems it enters. If our organisation’s learning system is rigid or fearful, technology will magnify that friction. If it’s curious, adaptive, and relational, AI becomes a force for exponential growth.

The future of work won’t be defined by artificial intelligence alone, but by learning intelligence - the shared capacity of humans and machines to learn together in service of meaningful purpose.

As CFOs and Finance Directors, we have a choice: continue funding AI as technology projects and accept mediocre returns or recognise AI adoption as a learning challenge and invest accordingly.

The leaders who win this era will be those who treat learning as infrastructure - as critical to AI strategy as data, capital, and talent.

Three Actions to Begin

If this resonates, here are three practical steps to start building learning power in your organisation:

First, commission a learning power diagnostic across teams involved in AI adoption. Understand where curiosity, agency, collaboration, or sensemaking may be low. Tools like WILD Learning’s assessment provide baseline data and ongoing tracking.

Second, include learning infrastructure in your next AI business case. Cost the time, facilitation, coaching, and reflective spaces required to build adaptive capability. Make it as visible as technology costs.

Third, add learning indicators to your transformation dashboard. Track them with the same rigour as financial metrics. When learning power rises, AI adoption follows.

An Invitation

The real competitive advantage in AI isn’t speed of deployment - it’s depth of integration. That integration depends entirely on our people’s capacity to learn through disruption.

As finance leaders, we’re uniquely positioned to champion this shift. We control resource allocation. We shape strategic priorities. We model what matters through our own behaviour.

Let’s use that influence to build organisations that don’t just implement AI but metabolise it - turning disruption into sustained competitive advantage through the power of collective learning.

The future belongs not to the most technically advanced organisations, but to the most learning-powered ones.

And that future is built on investments we approve today.

This article draws on research by Professor Ruth Crick and colleagues at the University of Bristol and WILD Learning, spanning more than 25 years of studying the architecture of human learning and its application to organisational transformation.

A sense of belonging and the opportunity to think and learn together are the crucial foundations that develop organisational resilience. These competencies require deep listening skills, psychological safety and a shared language for learning and creating strategy together.

But what is it that enables people to learn and for teams to learn together? To lean into risk, uncertainty and challenge and emerge from that process stronger and wiser?

It’s a complex mix of values, attitudes, skills, dispositions and self-stories and scientific research undertaken by Professor Ruth Crick and team at Bristol University found that this complex mix can be expressed as eight dimensions of “learning power”:

(1) Mindful Agency (self-leadership) is the most powerful, predicting (2) creativity, (3) curiosity, (4) sensemaking and (5) hope and optimism. Two relationship dimensions work together – (6) belonging and (7) collaboration. (8) Orientation to learning is about a person’s openness to the unknown and their willingness to pursue reflective change.

The dimensions are not simply skills or competencies but holistic attributes that we all possess, all of which involve thinking, feeling and doing.

A robust, reliable and valid measurement model was subsequently developed, the purpose of which is to stimulate ‘self-directed change’.

Feedback about a person’s Learning Power dimensions enables them to pay attention to their world in two different ways:

· first to their own story and context and their sense of purpose

· secondly to the systems thinking skills they can operationalise to help them achieve their personal or professional goals.

The WILD team further developed a framework to “scaffold” individual and team “learning journeys”. Each learning journey follows a time-limited process which:

• begins with a statement of purpose which is set in a context or place

• develops a plan that is based upon self-reflection from the learning power diagnostic feedback

• journals actions taken towards realising the plan

• ends in a review which includes a second diagnostic profile, providing clear data on impact and stories of significant change. This data can be analysed at individual, team and organisational scale.

As participants learn about themselves and how they typically ‘lean in’ to complexity, they experience meta learning – they “learn how to learn”. As they problem-solve, they naturally come to understand themselves as learners and make more productive decisions as they progress their projects. At the same time, they experience using systems thinking skills and strategies that lead to more resilient personal and professional outcomes.

Each individual or team learning journey is “scaffolded” – that is it follows an inquiry cycle, using both systems thinking skills and Learning Power to tackle complex problems in a way that results in more sustainable and robust outcomes.

WILD’s Learning Power app uses analytics to provide unique personal intelligence alongside coaching prompts and improvement ideas to help individuals build new knowledge and create value.

A WILD Learning Power profile is created through a 62-question proprietary assessment, and results are displayed against the eight dimensions of learning represented as a spider diagram. Coaching prompts, guidance and resources help individuals to become more self-aware to improve, account for and take responsibility for, their own learning journeys.

Backed by science and research, the Learning Power self-assessment is subject to the rigorous quality of social science research in terms of reliability, validity, trustworthiness and authenticity with over 100,000 data points and 24 years of research.

The diagram below shows three Learning Power profiles on a continuum of self-leadership

WILD also offers an Accredited Practitioners Programme to support the development of the meta-skills of self-leadership, learning relationships and thinking skills for complex problem solving.

This will equip participants to facilitate these meta-skills with others through:

· Coaching for Learning with individuals and teams using WILD’s research-validated tools and templates.

· Facilitating WILD Learning Journeys with clients (internal or external) focused on context-driven behaviour change and innovation in their own context, at individual, team or community levels, contributing to performance improvement.

This article draws on research by Professor Ruth Crick and colleagues at the University of Bristol and WILD Learning, spanning more than 25 years of studying the architecture of human learning and its application to organisational transformation.