In August 2023, many Salesforce teams were being asked for an AI roadmap before they had a dependable definition of a customer. The pressure was understandable: generative AI had changed executive expectations in a matter of months. But the fastest path to a useful pilot was not another prototype. It was a data contract.

A data contract is a short operating agreement covering meaning, ownership, quality and permitted use. It makes the invisible assumptions inside a demo explicit before those assumptions become production risk.

Cloud Group point of view

AI does not need every enterprise record to be perfect. It does need a deliberately bounded source of truth. The right question is not “Is our data clean?” It is “Which decisions will this AI support, and what evidence is safe and sufficient for those decisions?”

A practical playbook

The strongest next step is narrow enough to govern and useful enough to produce evidence. We would structure the work around these moves:

  1. Name the business decision or action before naming the model.
  2. Define the five to ten fields that determine identity, eligibility and context.
  3. Assign one accountable owner for meaning and one for technical reliability.
  4. Document freshness, consent, retention and the conditions that require a human handoff.
  5. Create a small evaluation set from real, approved customer scenarios.

The architecture and operating implication

Use Salesforce Data Cloud only where unification or activation creates a material advantage. Keep source-system authority clear, make calculated insights explainable and avoid copying data simply because a connector exists. A lightweight identity and access design will outperform an ambitious customer-360 program that cannot be governed.

Measure what changes

Model activity is not a business result. Track a small set of indicators that connect behavior to accountable work:

  • Match confidence and unresolved identity rate
  • Percentage of AI answers supported by approved evidence
  • Time from source change to usable customer context
  • Human correction rate by use case

The data contract is not paperwork before innovation. It is the mechanism that makes a fast pilot credible, a production launch governable and the next use case cheaper than the first.

Primary sources

This field note is grounded in the product and market context available at the time of publication.