AfrexAI

    Free Tool

    AI Use Case Scorecard

    Eight questions that decide whether a workflow is worth automating. Answer them about one specific process, not about your business as a whole.

    Nothing is sent anywhere. Your answers stay in this browser tab and are gone when you close it.

    0 of 8 answered

    1. 1. How often does this work happen?

      Automation earns its cost through repetition. Work that happens twice a year rarely pays back a build.

    2. 2. How consistent are the inputs?

      Consistent inputs make the exceptions countable. Inputs that vary every time turn the build into open-ended discovery.

    3. 3. Does one person own this process today?

      An owner defines the exceptions, approves the new version, and is accountable for adoption. Processes with no owner stall in review.

    4. 4. Can you put a number on what it costs now?

      Without a baseline there is nothing to compare the result against, and no way to tell whether the build was worth it.

    5. 5. What happens if the output is wrong?

      Error tolerance decides how much human review the design needs, which changes both the cost and the achievable time saving.

    6. 6. Can you get programmatic access to the systems involved?

      Every integration needs an owner, a credential, and a rate limit. Systems with no API and no export are the most common hidden blocker.

    7. 7. Does the information needed already exist in a findable place?

      A system cannot use knowledge that nobody has written down. Undocumented judgement has to be captured before it can be automated.

    8. 8. Will the team using it want it?

      A system the team works around is a system that failed. Adoption is decided by whether the new process is easier than the old one.

    Answer all 8 questions to see the result. Partial answers are not scored, because a single unanswered question is usually the one that decides the outcome.

    How the scoring works

    Each question scores 2, 1, or 0, for a maximum of 16. The eight criteria are weighted equally on purpose. In practice any single zero can stop a project on its own, which is why the result calls out zero-scoring answers separately from the total.

    The criteria come from the conditions that separate AI projects which reach production from those that stall: repetition, input consistency, ownership, a measured baseline, error tolerance, integration access, available knowledge, and team demand. Published research on stalled projects points at the same causes, mainly brittle workflows, integration complexity, and misalignment with day-to-day operations rather than model quality.

    A low score is a useful answer. It usually means the process work has not been done yet, and that work is worth doing whether or not you ever automate the workflow.

    Questions about the scorecard

    What makes a good first AI use case?

    Recurring work with consistent inputs, a named owner, a measurable current cost, tolerable error consequences, and programmatic access to the systems involved. Those five conditions matter more than which model or platform you choose.

    Does a low score mean AI cannot help us?

    No. It usually means this particular workflow is not the right place to start. Low scores commonly point at missing process documentation or an absent owner, both of which are worth fixing regardless of whether you ever automate the work.

    Why does error tolerance change the design?

    Work with serious consequences for a wrong answer needs a person in the loop on the decisions that carry the risk. That is a legitimate design, but it changes the achievable time saving, so it should be priced in at the start rather than discovered later.

    Is this scorecard a substitute for discovery?

    It is a filter, not an assessment. It tells you whether a workflow is worth examining properly. Real discovery measures the current cost, maps the exceptions, checks the integrations, and agrees the success measure.

    Cost CalculatorAI Implementation Services