Andrew Harrison

Andrew Harrison

Agentic & GenAI Governance Advisor ABN AMRO
Andrew Harrison

Andrew Harrison’s career has focused on the development and control of complex, innovative, technology-driven systems in Financial Services, Technology, and Public Sectors. He has worked with Deloitte, HSBC, Thomson Reuters F&R GRC (bought by LSEG), and startups, working professionally across the UK, Germany, and the Netherlands. He currently serves as Agentic & GenAI Governance Advisor at ABN AMRO Bank N.V. in the Netherlands, having previously been an AI Engineer and AI Analytics Lead in Strategy & Innovation. His focus has been on controlled acceleration of GenAI within the bank, through governance upgrades, designing technical working groups, serving on the GenAI use case board, providing subject matter expertise, and practically implementing the EU AI Act. Currently he is setting up the Agentic AI governance framework for the bank. Andrew has a multidisciplinary background with Bachelors in Psychology, Business, and Computer Science, and Masters in Occupational Psychology, International Relations and Artificial Intelligence. He marries deep technical and analytical understanding, with expertise in operational process design, and complex systems thinking. He is deeply passionate about all aspects of AI; technical developments, governance, law, risk management, policy, existential safety, and security.

Tuesday 14th April - Conference Day One

3:40 PM Closing Panel Discussion - Security & Abuse Guardrails for Autonomous Actions

Agentic AI doesn't just automate tasks - it makes decisions, executes transactions, and acts on behalf of customers at speed and scale. That's a competitive advantage. It's also a largely unmapped attack surface.
This session examines what happens when autonomous agents go wrong - through adversarial manipulation, unexpected failure modes, or actions that fall outside the boundaries anyone thought to define. The panel moves beyond threat awareness to what effective guardrails look like in production: how they're designed, where they break down, and who owns them when they do.
  • Mapping the agentic abuse surface: prompt injection, credential misuse, and actions that are technically permitted but commercially damaging
  • Designing controls that constrain agent behaviour without killing the autonomy that made deployment worthwhile
  • Establishing accountability when an agent - not a human - initiates a transaction that goes wrong

Check out the incredible speaker line-up to see who will be joining Andrew.

Download The Latest Agenda