Teaching Your AI Coding Agent What It's Allowed to Build-Financial Services View

Talk;

Teaching Your AI Coding Agent What It's Allowed to Build - A Financial Services View

Most companies enforce architecture standards the old way: a review board and a checklist before anyone writes a line of code, then a wall of blunt, late guardrails once they actually start building. You find out you broke a rule when the deploy fails with no explanation, or worse, when something just quietly gets deleted overnight because it didn't comply. Nobody actually learns the policy, they just learn where the walls are.

This talk shows a different way of doing it. Instead of gatekeeping after the fact, the governance lives inside the AI agent's own reasoning: it reads your actual policy docs and tells the developer what applies, in plain English, while they're still describing what they want to build, before any Terraform exists. Then, as a backstop, the same policies get mirrored into OPA/Rego rules that the agent checks itself against before it even shows you a design, and that CI checks again if anyone tries to sneak something past it by hand.

• Agenda;
  18:00; Doors Open
  18:30; Speaker delivers talk
  19:15; Pizza and catch up
  20:00; Doors close

Location; CGI Fothergill House, 16 King St, Nottingham NG1 2AS