Scoring Process Candidates

Definition. Process scoring is a structured comparison of possible AI projects before a team commits to a pilot.

The most visible process is not always the best place to start. A team may see substantial manual work yet lack a way to measure the result, access the right information or put the AI output into daily use. A short, consistent comparison makes those gaps visible while there is still time to choose another process.

Check readiness before assigning scores

For each candidate, ask three questions:

1. What result would improve? Name an outcome the business already recognises, such as a request resolved, a record accepted or an order fulfilled. Establish how it is measured now and what quality must be preserved.
2. Is the information available? Identify the records, rules and examples the system would need. Check whether the team can use them and whether they represent the work arriving today. Data that exists somewhere but cannot be accessed or trusted is not ready for a pilot.
3. Where will the output go? Name the person or system that receives it, the action that follows and the point at which a human reviews it. A good model answer sitting in a shared folder does not improve a process.

If a candidate fails one of these checks, record the missing work. It may be a good opportunity later, but a high projected benefit does not make its data or workflow appear.

Compare the candidates

Use the same criteria for every process that passes the readiness check. A simple scorecard can cover:

If you use weights or a numerical scale, agree on them before seeing the totals. Show the evidence behind each rating and mark assumptions that a pilot must test. A score is a way to compare reasoning, not proof that one process will deliver a return.

Some conditions should stop a launch regardless of the total: no accountable owner, no viable path to use the output, or a material risk that has not been addressed. Decide what those conditions are with the people responsible for the process and its controls. Do not hide them inside an average score.

Make a decision the team can revisit

Choose a small number of candidates for further design and say why the others were deferred. For the leading candidate, write down the outcome definition, current baseline, main uncertainty and the evidence needed to continue after a pilot. That turns selection into a testable decision rather than a promise.

The process owner should be accountable for the shortlist. Finance, operations, engineering and relevant control functions can challenge the estimates from their own perspectives. If seniority determines the result before the comparison begins, the scorecard does not inform the decision.

Scoring is useful when there is a real choice. If a requirement already dictates the work, use the same questions to plan and control it, but do not present the exercise as a competition between projects.