Approach

Make the investment case
before the build.

Start with the work. Compare alternatives. Keep forecasts separate from measured results.

Follow the evidence.
Not the momentum.

A pilot is a way to answer a business question, not a commitment to scale. Establish the baseline, run a bounded test, then make the decision.

  1. 01

    Establish the baseline

    What happens today?

    • Map one process and its exceptions.
    • Measure time, quality and full cost.
    • Compare AI with simpler alternatives.
    Evidence to carry forwardA clear starting point and a testable value hypothesis.
  2. 02

    Run a bounded pilot

    Does the change help?

    • Define success, failure and ownership.
    • Test representative work and exceptions.
    • Track quality and actual human effort.
    Evidence to carry forwardObserved results, with the scope and limits made clear.
  3. 03

    Make the decision

    Is it worth continuing?

    • Compare results with the baseline.
    • Include operating and change costs.
    • Check risk, ownership and readiness.
    Decision, not a defaultImprove, expand or stop.

    Not using AI can be a useful result.

A measurement plan, not claimed results. This is a proposed working approach, not a claim of measured client outcomes.

What counts
as value?

Measure the process, not just the model. Use the same definitions before and during the pilot, and compare with the simplest credible alternative.

01 Time
How long does the whole task take, including waiting and human review?
02 Quality
Does the work meet the agreed standard, including difficult cases?
03 Rework
How much correction, exception handling and repeated work remains?
04 Full cost
What does a successful outcome cost, including people, systems and ongoing operation?

What’s getting in the way of better work?

Tell me about the process, the outcome you need and the constraints you face.

Please don’t send confidential business or customer information.