You Don't Need a Model of Your Board
Why executives should train their own read of the room rather than outsource stakeholder prediction to AI, and the augmentation test for judging any decision-support tool.
About
About me
I'm a decision scientist. I finish an Executive MSc in Behavioural Science at the London School of Economics in December 2026, and my work sits between behavioural science and the design of AI systems.
I run Behaviorist, which works on the points where an AI system is allowed to act on its own and where it has to stop and ask a person. Before that I built PeopleNotResumes, which looks at how candidates actually make decisions rather than what their CV claims.
I write here about one problem: what pressure does to judgment, and what happens when a machine is standing by to help.
You now think with three systems. Only two of them are yours.
- System 1Your gut
- The instant read. You walk into a room and know something is off before anyone says a word. It's fast, it's free, and it's right often enough that you stop checking it.
- System 2Your reasoning
- The part that slows down, lays out the options and changes its mind. It burns real energy, and your body only pays for it when nothing more urgent is happening.
- System 3Artificial cognition
- System 3 is artificial cognition that operates outside the human brain, as an external layer of decision-making. It extends Daniel Kahneman's dual-process theory of human thought: System 1, fast intuition, and System 2, slow deliberation. Checking its output is how your two human systems stay engaged, and it is hardest exactly when you most want the answer.
Try it
Pressure decides which of your systems you can reach. Drag the slider.
CalmUnder pressureOverloaded
System 1Always on.On
System 2Working. Slow enough to change your mind.On
System 3 handoffYou ask it for options, and you actually read them.Holding
You're making a decision.
Who I work with
Leaders
Decision quality under pressure, for people whose calls carry weight, and who now make them alongside AI.
Organisations
Through Behaviorist, the human side of agentic AI: when a person should step in, whether they are actually thinking, and whether their decision changes the system.
Researchers
Open questions on human judgement alongside machine cognition, starting with the Behaviorist lab's research agenda.
Work
What I do
Two practices and one advisory role, held together by one question: where a decision should sit with a person, and where it can safely sit with a system.
BehavioristHuman-agent decision architecture
Decision architecture for agentic AI systems: where an agent is allowed to act on its own, and where it has to stop and ask a person.
- Audit — map every point where the system decides without a human.
- Design — set the checkpoints, escalation rules and evidence trail.
- Checkpoint Copilot — the running layer that holds those checkpoints in production.
PeopleNotResumesBehavioural hiring intelligence
Hiring judged on how a candidate actually decides under pressure, rather than on what a CV claims.
Compassion LabsAdvisor
A nonprofit teaching metacognitive skill to executives and workplace leaders inside organisations. Attention and breath practices first, then noticing thought patterns as they arise, then applying that to real decisions under pressure. Programs run across the US and UK and are measured with the CAWI Index, their Compassion at Work Inventory.
- Practice, attention and breath, six weeks.
- Awareness, naming reactions before they steer you, eight weeks.
- Leadership Application, judgment under pressure, twelve weeks plus coaching.
Their premise sits next to mine. AI now handles much of the drafting and first-pass analysis, so producing thinking is no longer the bottleneck. Judging it is.
Where it breaks
Under pressure the slow part goes first. The gut stays, and now it has a research assistant. It asks for a memo and gets one: confident, well sourced, and built around the answer it already had. Nobody in the room experiences this as bad judgment. It feels like moving fast.
Which is why measuring people doesn't fix it. Nobody on your team got worse at thinking. They lost access to it.
What is established, and what is not
- Breathing at roughly six breaths a minute measurably settles the nervous system, and it has been tested properly. That part isn't controversial.
- Polyvagal theory gets the credit for this almost everywhere. Its actual anatomy is disputed. I use the breathing and the measurements, not the theory's name.
- Stress narrowing how people think is well established. What isn't established yet is my part: that a stressed person using AI ends up confirming what they already believed. That's the claim I'm testing.
Research
Research & writing
What Actually Happens When System 2 Goes Offline
The physiological mechanism behind executive judgment failing under pressure, why heart rate variability tracks it, and how AI changes the failure mode rather than fixing it.
The Board Doesn't Have One Nervous System
Why a boardroom's collective state, not the merits of the proposal, often decides the vote, and how a confident AI output becomes the easiest way for a dysregulated room to avoid its own disagreement.
System 3 Is Artificial Cognition
System 3 is artificial cognition outside the human brain. How Systems 1 and 2 stay engaged when the machine supplies an answer, so it gets evaluated instead of absorbed.
Contact
Work with me
- Decision scientist
- Executive MSc Behavioural Science, LSE (Dec 2026)
- Founder, Behaviorist
I work with leaders, organisations and researchers on how people think alongside AI. Write to me and tell me what question or decision is on the table.