A useful starting point is a specific piece of work, a measurable outcome, and someone who can change the process.
Start with the work
Name the task before choosing the technology. Who performs it, what information do they need, and where does the process slow down? A request to “use AI in operations” is difficult to evaluate. A proposal to help a service team prepare a first response from approved guidance has a clearer boundary. Observe the current workflow and include the people who handle its exceptions.
Compare value with difficulty
Consider the time spent, the frequency of the task, the quality required, and the cost of a mistake. Then examine the practical constraints: access to information, integration effort, review capacity, and ownership. A frequent task is not automatically a good candidate if every result needs extensive correction. Prefer an opportunity where the expected benefit can be tested without committing the whole organisation.
Define a decision at the end
A pilot should help you decide whether to expand, change direction, or stop. Set an initial baseline and agree on the evidence you need. For the service team, that might include response preparation time, factual accuracy, and the amount of editing required. Record what would make the project unsuitable, as well as what would justify continuing.
Give it an owner
Choose someone responsible for the outcome and able to improve the surrounding process. Set aside time for users to try the system and report problems. The first project should leave you with a better understanding of the work, even if the proposed AI solution does not progress.
