Agentic AI
An agent acts. A role still owns the step.
An AI agent plans, uses tools and acts until a goal is met or a limit stops it. That moves the risk: from what the model says, to what it is allowed to do.
What changes
A chatbot answers. An agent takes the next step: it searches, drafts, files or sends, then looks at the result and continues. The loop is only as safe as the limit around it, and the person who can stop it.
Role paths already ask a team to name who checks, who decides and who fixes. An agent does not replace that agreement. It makes the missing name show up faster.
Work through it in this order
-
Why AI tools aren't enough
An agent can take a step. It still needs someone who checks, decides and fixes.
Open → -
Your agentic self
Each person designs an agent that extends their role, inside one team agreement.
Open → -
Where people step in
People can set limits, check, pause, correct or stop an agent at each step.
Open → -
Shape your role
Each person sees the problem in their role, and what a path and an agent change.
Open →
Start from the team, not the tool
The proposed pilot is the role paths. This section is how an agent fits inside them.