The pilot ran.
What would make you say no?

Testing AI is an activity. Deciding whether a particular process deserves further investment is a commercial question. A pilot should help answer the second.

I would distrust a review where a working demo, a positive comment and a completed booking can all be presented as the same proof of success.

Specify the problem the pilot should reduce

Start with the difficulty in the existing journey. Is the intended change clearer qualification, a more useful booking route or another defined process improvement?

Then agree what a valid lead and a qualified booking mean for this customer. A supplier and customer using different definitions can finish the same pilot with incompatible conclusions.

A commercial pilot needs a reason to say no.

Compare groups that can support the decision

I would ask for the baseline and its limitations before interpreting improvement. Historical information may be incomplete, but that should be visible rather than silently ignored.

Keep lead source, working hours and customer segment in view. Mixing unlike groups can make a difference look like an effect of the pilot when the comparison does not support that conclusion.

Define stop, change and expand in advance

Agree what evidence would justify expansion, which finding would require a change and what would make the parties stop. Do that before seeing which result looks most flattering.

The decision need not be an arbitrary universal percentage. It needs to address the customer’s original problem and the assumptions on which the investment depends.

Keep failures and uncertainty discussable

A useful review should explain missing information and unsuccessful cases alongside the positive outcomes. Uncertainty is a reason to qualify the conclusion, not to rename the objective afterwards.

For Conversion24, I would want the customer to leave with a clearer next investment decision. The pilot’s purpose is not to prove that the supplier was right.

Agree the decision before the result gives everyone a reason to reinterpret it.

Nick BlomFounder, Conversion24Published 11 November 2026

An opinion article by Nick Blom. The examples illustrate process choices and do not establish measured Conversion24 results.