Our very own Sarah Burrows took the stage at the GAT AI Sprint panel on 3rd November to discuss "Trustees in the ageof AI" alongside industry peers. The consensus? AI can draft a brilliant trustee paper, but it absolutely cannot carry the fiduciary responsibility for the final decision. 

One of the most critical discussions centred around governance and the very real risk of "agent sprawl" where well-meaning staff can easily create hundreds of similar custom Copilot agents that all produce completely different outputs. The panel also highlighted a hidden danger: underlying models are constantly updating, and those updates fundamentally change how your custom agents reason. While Copilot might flag an update, managing the version control and rigorously re-testing those custom agents rests entirely with the Board.

As highlighted during the governance discussion, firms need to govern AI agents the exact same way they govern their people. You wouldn't let an untrained junior make a massive capital distribution decision unchecked, so why let a general-purpose chatbot do it?

If your firm is moving beyond basic experimentation, Aspida is here to keep your digital workforce in line and we can help with:

• AI Governance Frameworks & Guardrails
• Generative AI Policies, Procedures and Controls
• Board-Level Accountability & Reporting
• Agent Risk Assessments & Testing
• Staff Training & Adoption

Don't let an unexpected model update become your next compliance headache. Reach out to see how Aspida can help you build a robust frameworkbefore you deploy your next AI tool.