Agentic AI

For teams past chat, moving to systems that take actions — and the discipline that requires.

Technical leads, ops owners, and anyone who will be responsible when an agent does something unexpected.

60–90 minutes Your office or remote Leadership + team No deck — live systems
  • One agent running against a real internal task
  • An evaluation harness for it
  • A written approval and escalation model
01

Agents versus workflows

When a deterministic pipeline is the better answer, which is more often than the market admits.

02

Tools, memory, and context

What an agent needs to be useful, and what happens when it is missing.

03

Evaluation

Knowing whether it works before your customers find out.

04

Failure and blast radius

Sandboxes, approvals, and where the human checkpoint belongs.

← All sessions