This talk examines what changes once an AI agent can do more than generate text: when it can access tools, investigate systems and potentially trigger consequential actions. It addresses how to structure read, recommend and execute permissions, where to place human approval points, and how to apply scoped credentials and revoke them. It also covers how to leave a useful evidence trail across an agent workflow.
Speaker - Hammed Ajibade (https://www.linkedin.com/in/hammed-ajibade-010bb2116/)
Hammed Ajibade is a Lead Software Engineer at Pulse. He builds and leads software across payment systems, platform engineering, mobile products and practical AI tooling. His work sits close to money movement, where systems have to stay reliable under real use.