A useful pilot answers a question that matters to the next investment. For example, can the team isolate one booking rule while keeping the existing appointment record and confirmation behavior correct? Define what the team expects to observe, which representative cases it will use and who can judge the result.
A pilot can also fail to answer the larger question. A demonstration using artificial data may show that a screen works without revealing how real scheduling conflicts, notifications or support exceptions would be handled.
Count the temporary work needed to make the pilot coexist with the old flow. Who owns the data during the trial? How will a mismatch be detected, and what happens if a customer changes an appointment after the new path has been used? If the team cannot safely isolate the rule, that finding may change the scope before development begins.
A small first deliverable is not automatically a small total investment. Compare its integration, verification and removal work with the work required to keep the present system operating.