Can you name every AI agent in your firm?
The FCA's AI good practice publication treats knowing what AI you run, and who owns it, as a foundation of governance. Here's what it calls good and poor practice.
Can you name every AI agent in your firm? It sounds like a simple question. For many firms it isn't, because AI now arrives through vendors, internal teams and individual staff, often faster than anyone can track.
The FCA's AI good practice publication treats knowing what you run as a foundation of good governance.
What the FCA calls good practice:
💡 A central inventory of AI agents, showing who owns each one and what data and systems it can access.
💡 A clear inventory of AI use cases and the models they rely on, separating experimental tools from production ones.
💡 Every material AI use case mapped to a named senior manager, as one smaller firm in the publication did
What it calls poor practice:
❗ "Unclear responsibility for AI systems", especially when ownership is split between technology and business teams.
❗ "Gaps between AI risk assessment and SMF accountability", where identified risks have no clear owner.
You can't govern an agent you don't know exists, and you can't hold someone accountable for an agent nobody owns.
📖 Read the FCA publication