Lesson 04 · Sales & delivery

Sell a first step you can finish

A paid pilot should answer a decision. Define one process, a group of users, the information you can use and the actions allowed. Agree the baseline, fee and evidence needed for a wider deployment. The customer can then make a manageable first commitment, and the partner has a clear piece of work to price.

Choose representative questions and expected answers before testing. Include missing information, older versions and requests that cross access boundaries. For a shipment assistant, showing another customer’s record may be more serious than failing to answer a question. The consequences of an error determine which failures stop rollout.

Measure the complete task. Time spent checking, correcting and handling exceptions belongs in the result. If a system produces a draft quickly but moves work into a review queue, the whole task may not have improved. Record the tested configuration and known limits. A successful pilot describes that setup and those conditions; it does not prove that every future version will work.

Check your reasoning.

Choose one answer for each question. You can retry; the explanation is part of the lesson.

1. Which is a useful pilot acceptance condition?
2. An assistant drafts in two minutes but requires fifteen minutes of review. What should you measure?
Open the answer key

1. Representative tasks meet agreed criteria, including important failure cases.
Acceptance needs task-specific evidence. A polished demonstration or general benchmark cannot establish the customer’s required outcome.

2. The complete task, including review, correction and exceptions.
Task-level effort is what the customer and partner must fund. Moving work to another person does not remove its cost.