Engineering Control into Agentic AI That Controls Itself [RSVP REQUIRED]

This event requires an RSVP; to attend, submit form here

Agentic AI changes the problem of trust from a policy question into an engineering one, and it changes the ambition from automating tasks to reinventing how work gets done.

This roundtable examines how application-layer design choices determine whether autonomous systems remain controllable in high-risk environments, covering:


1.Agent boundaries

2.Permissions and guardrails

3.Memory and context engineering

4.Tool use

5.Loop engineering

6.Human-in-the-loop controls


We'll explore how control and risk management can be engineered directly into agent architectures, built on ATR (Auditable, Traceable, Repeatable) properties, rather than enforced through after-the-fact reviews. When agents can be trusted with real execution, the question stops being "where can I add AI" and becomes "how should this function be redesigned." Participants will exchange practical patterns, trade-offs, and hard-won lessons from building agentic systems that must be auditable, resilient, and aligned by design, and from rethinking the workflows those systems now make possible.

This roundtable is for builders and technical leaders who want to engineer trust in, not bolt it on, and use that trust to reinvent the workplace, not just automate it.