A three-day residential executive programme for board members and senior managers who must allocate capital, govern AI deployments, and make consequential decisions in firms where AI is now a material input.
The programme is taught by a Cambridge economist and machine learning practitioner, supported by a senior industry practitioner and a board-level guest. It is held in residence at the Møller Centre at Churchill College — two nights' accommodation, a college dinner, and a working cohort of 25–30 senior leaders from deliberately mixed sectors.
The programme's promise is specific: delegates return with five concrete artefacts — each tied to a real decision in their own organisation — that are auditable, presentable to the next level up, and usable on Monday morning.
Board members (executive and non-executive), C-suite leaders, senior managers being groomed for board-level roles. CFOs, COOs, Chief Risk Officers, Chief Strategy Officers, and directors sitting on AI, technology or audit committees.
Three-day residential at Møller Centre, Churchill College, University of Cambridge. Two nights' accommodation, college dinner, cohort of 25–30. No technical prerequisite. Pre-course online module included.
£6,500 per delegate, all-in residential. Includes all accommodation, meals, venue, course materials, pre-course module, and post-course alumni engagement. Target launch: Autumn 2026.
Most executive AI programmes are organised around the technology. This programme is organised around the decision — every session asks how AI changes a particular kind of executive choice.
The organising framework is Judea Pearl's Ladder of Causation, taught as a working tool rather than a philosophical curiosity. Every session maps explicitly onto a rung of the ladder, and every take-away artefact requires the delegate to climb it — shifting senior leaders from level-1 thinking ("the data shows…") to level-2 and level-3 thinking.
"What is correlated with what?" Most dashboards and most AI predictions live here. Useful for monitoring; insufficient for action.
"What happens if we do X?" The level at which decisions actually live. Requires either experiments or a causal model.
"What would have happened if we had done Y instead?" Modern AI makes this practical at scale for the first time.
"What will happen?" — demand forecasts, churn risk, default probability. AI is mature here; the risk is overconfidence in models trained on stable past distributions.
"What will happen if we do X?" — the question every board paper implicitly asks. Where most AI deployments fail in their second year. The central intellectual differentiator of this programme.
"How should we present, draft, summarise, brief?" — the productivity layer where LLMs are most useful, and where the largest near-term gains and the largest reputational risks both sit.
"Should this decision be automated?" — the new question. Agentic systems can take action, not just provide advice. Requires governance: checkpoints, audit trails, reversibility.
Day 1 calibrates. By the end of the day every delegate shares the same mental model of how AI works, where productivity gains actually come from, and how to sort the decisions in their portfolio by the type of AI tool that fits.
Session 1 draws on the 86-slide Learning Machine deck (Festival of Ideas, March 2026) — the four-city historical arc (Florence 1490, Scotland 1776, Flint 1985, Cambridge 2026), the Brynjolfsson J-curve, the Acemoglu optimist–pessimist GDP range, and the Three Tiers of Knowledge. Session 3 builds the AI Decision-Mode Selector from real cases submitted in each delegate's pre-course diagnostic.
Day 2 is the intellectual heart of the programme and its central differentiator. No peer programme in the executive AI market explicitly teaches causal reasoning to senior decision-makers.
Why most "data-driven decisions" are pattern-matching, why that breaks under intervention, and what to do instead — using AI to reach the higher rungs of Pearl's Ladder rather than to dress up associational thinking in a dashboard.
The Causal Decision Canvas — Session 3's signature artefact — forces each delegate to name a real decision, draw its causal graph, identify confounders, specify AI's role, pre-commit to a decision rule, and state what would change their mind. Faculty rotate among trios to challenge graphs and rules. By end of session every delegate holds an audit-quality Canvas for one live decision.
Day 3 is the operational day. Every delegate builds a working agent tied to a real workflow in their organisation, completes the governance memo, and presents a 90-day decision plan to peers for public challenge.
The agent build lab pairs each delegate with a teaching assistant. Over 90 minutes they configure an agent — Claude with MCP, custom Skills and a knowledge file — that automates one component of the decision mapped on their Day 2 Canvas. The deliverable is genuinely working: at close of session each delegate's agent runs and produces output they can show colleagues on the journey home.
Senior executives do not buy three days at Cambridge for a tour of generative AI mechanics. Each artefact is built during a specific session, tied to a decision in the delegate's own organisation, and designed to be auditable, presentable to the level above, and usable on Monday morning.
One-page flow diagram routing any decision to the right AI tool: prediction, causal estimation, generative drafting, or agentic automation. Built Day 1, Session 3.
The programme's signature artefact. One-page diagnostic: name the decision, draw the causal graph, identify confounders, specify AI's role, pre-commit to a decision rule. Built Day 2, Session 3.
Two-page critique applying Pearl's Ladder to a real AI vendor productivity claim — associational or causal? What evidence would upgrade it? Built Day 2, Session 2.
A functional AI agent tied to a workflow in the delegate's firm, plus a one-page governance memo specifying checkpoints, audit trail, reversibility, and failure escalation. Built Day 3, Sessions 1–2.
Structured one-page format for presenting a decision including the "what would have happened if…" scenarios AI helped generate. Disciplines uncertainty into the briefing. Built Day 2, Session 4.
One decision to reframe with AI, one agent to deploy, one process to redesign — with milestones, a sponsor, and 30/60/90-day check-ins committed to peers in public. Built Day 3, Session 4.
★ The Causal Decision Canvas and the Working Agent + Governance Memo are the programme's signature outputs — the intellectual and visceral centrepieces respectively.
The programme is led by a Cambridge economist and ML practitioner, supported by a senior industry practitioner and board-level guest speakers.
Dr. Melvyn Weeks
UNIVERSITY OF CAMBRIDGE
Cambridge economist with expertise across machine learning, causal ML, generative AI, and agentic AI. Teaches the Cambridge MPhil course in Causal Inference and Machine Learning (DS300). Principal lecturer for Day 1 Session 1, all of Day 2, and Day 3 Session 3.
To Be Confirmed
INDUSTRY
A serving or recently serving AI deployment lead from a major bank, retailer, industrial firm, or consultancy. Co-leads the agent build lab (Day 3) and the vendor-claim memo session (Day 2). The highest-leverage faculty decision for the programme.
A non-executive director who has chaired an AI or technology committee at a listed firm. Format: 30-minute interview by the Programme Director, 30-minute Q&A. Topic: "What I wish my board had asked me."
A guest from Anthropic, OpenAI, Google DeepMind, or a comparable frontier lab speaking on model trajectories. Used selectively when a particular cohort would benefit.
£6,500 per delegate, all-in residential. Two open-enrolment cohorts planned per year — Autumn 2026 and Spring 2027. Bespoke single-company variant available.
£6,500 per delegate. Includes two nights at Møller Centre, all meals, college dinner, course materials, pre-course module, and post-course alumni engagement. Cohort of 25–30.
Many buyers send 2–3 delegates — £13,000–£19,500 of corporate spend, typically within senior leaders' personal sign-off authority. Group discounts available on request.
Same intellectual architecture, customised to a single client's vendor stack, sector and board agenda. Typical range: £150,000–£200,000 per engagement for a full cohort at Møller.
A board team of three, three days at Cambridge, under £20,000.
Recommended framing for corporate purchasersStructured email check-in at Day 30. Small-group alumni call at Day 60. Full-cohort reunion call at Day 90 — each delegate presents what changed in their decision-making practice.
Cambridge AI Fellows reunion at Møller Centre. Optional but cohorts attend at high rates. Alumni network is part of the £6,500 value proposition.