How should a model lab narrow its first customer change pilot?

Begin with a capability question that the lab cannot answer from its current evaluations. A useful question is whether the model recognises when a customer's changed circumstances invalidate an earlier decision. That is narrower and more testable than asking for a complete business simulation.

Define the starting state, the customer-triggered change and the outcome that will demonstrate the capability. Then identify the minimum roles, records and rules needed to connect them. If a downstream process has no bearing on the decision, it can remain outside the pilot, with that boundary documented.

A buyer can ask MirrorFirm to scope a generated sample around this boundary and explain which variations it represents. The sample request should state the intended training or evaluation use, the model's available actions and the evidence the lab expects to inspect.

A useful first step is to write a brief containing a capability hypothesis, an observable failure and a proposed acceptance test. This gives both parties something concrete to assess before broadening coverage. A narrow pilot can support a clear conclusion about the selected capability; it should not be presented as proof of general business competence.