AI and technology executive building and evaluating agentic systems
Group CTO & Chief AI Officer at Cape.io · AI safety researcher · Author
I work where frontier AI meets real organisations: building agentic operating models, evaluating autonomous-system behaviour, and turning safety evidence into decisions that boards and engineering teams can use.
For more than 25 years, I have led technology, AI, data, and platform organisations across AdTech, financial services, and media. I am currently Group CTO and Chief AI Officer at Cape.io, a Research Associate with Arcadia Impact's AI Governance Taskforce, and a guest industry expert for MIT xPRO's AI programme for senior leaders.
Lessons from External Review of DeepMind's Scheming Inability Safety Case, presented as an oral at the Second Workshop on Technical AI Governance Research, ICML 2026.
We applied Assurance 2.0 to Google DeepMind's public safety case for Gemini 2.5 Pro, identified concerns affecting its scope and use in deployment decisions, and proposed a practical process for independent review.
Paper · ICML oral and video · OpenReview
AgentMisalignment: Measuring the Propensity for Misaligned Behaviour in LLM-Based Agents introduces realistic agent environments for evaluating shutdown resistance, oversight evasion, sandbagging, deception, and power-seeking.
Our experiments found that system-prompted personality can change misalignment propensity as much as, and sometimes more than, model choice. I co-developed the benchmark through the Cambridge AI Safety Hub.
Paper · Benchmark implementation
| Project | What it demonstrates |
|---|---|
| AgentMisalignment benchmark | Agent-based safety evaluations built with the UK AI Security Institute's Inspect framework |
| RSI Eval Lab | A runnable lineage auditor and Inspect task for evaluator tampering, shutdown resistance, containment, and Goodharting in recursively improving systems |
| AI Engineering Journey | A runnable maturity assessment that turns seven delivery capabilities into a phased transformation roadmap |
| Board AI Governance Toolkit | An inspectable AI risk register, decision templates, and deterministic board dashboard generator |
| Garmin Running Dashboard | A privacy-conscious Streamlit application that turns Garmin exports into training and performance analysis |
| Grow Your Business with AI: code samples | Practical examples accompanying my Springer Nature book |
| Personal research and writing | Interactive work on AI safety, AI engineering, governance, and agentic transformation |
- Group CTO and Chief AI Officer at Cape.io, designing an agentic operating model across engineering and business functions.
- Former Group CIO/CTO and AI/data leader at Fenestra, Experian, Kantar, GroupM, Havas, and Accenture.
- Contributor to the Bank of England and FCA AI Public-Private Forum, the EU AI Act industry process, and TechUK policy work.
- Guest industry expert teaching enterprise AI strategy, agentic organisations, and AI safety governance for senior leaders at MIT xPRO.
- Author of Autonomous Minds (Wiley, 2025) and Grow Your Business with AI (Springer Nature, 2023).
- How should evaluations adapt when an AI system can improve or redesign itself?
- What evidence should frontier developers provide before deployment?
- How can organisations scale agentic workflows without losing control, accountability, or engineering discipline?
I am based in Cambridge and work across executive leadership, AI safety research, board advisory, and applied AI collaboration.