The operating-model shift no one is staffing for

The work humans do up-levels almost overnight. Tier-1 cases become exceptions and edge cases. Tier-1 reps become exception handlers, judgment workers, and escalation owners. That's a different job description, a different hiring profile, and a different management cadence.

Most programs we benchmark have not redrawn the org chart. They ship the agent, watch attrition rise in tier-1, and only then realise the operating model was supposed to change too.

Every meaningful Agentforce program needs an agent product manager — someone who owns the agent's behaviour, prompt and skill catalogue, error rate, and roadmap, the way a SaaS PM owns a product. This is not the AI architect's job and it's not the line manager's job. It's a new role and most enterprises don't have it on the org chart yet.

Run the org redesign as a parallel workstream to the build. Identify which roles compress, which expand, and which are new. Name the agent PM before go-live. The cost of doing this after the fact is twice the cost of doing it up front.