Data Cloud isn't a project. It's a precondition.

Most enterprise programs we benchmark have Data Cloud and Agentforce as separate tracks, with separate steering committees and separate timelines. It feels like good governance. It is actually the most reliable way to push agent ROI from this fiscal year into next.

Agentforce only knows what Data Cloud has unified. If unified identity, profile, and case history aren't live when the agent ships, the agent ships with a partial view of the customer — and your service team notices in week two.

We don't mean finish Data Cloud before starting Agentforce. We mean unify the data the first agent needs before that agent goes live. Scope Data Cloud to the agent's use case, not to a five-year master data vision.

If the first agent is a service case-triage agent, you need unified profile, case history, and entitlements — not the full 360. Build the slice the agent will consume, ship it together, and expand the slice as you add agents.

'For the first agent we're building, which six data objects must be unified in Data Cloud before it goes live, and who owns each one?' If your delivery lead can't answer in a sentence per object, the program isn't ready to start.