All work

AI evaluation

Knowing when to stop a pilot.

Using evidence from a small retail pilot to decide whether an AI solution offered enough value under real operating constraints.

Context
Retail operations
My role
Use-case definition, coordination & assessment
Status
Pilot concluded; no further rollout

Starting point

  • Proposed AI use cases
  • Unproven operational fit

The change

  1. Small-site pilot
  2. Assessment of results
  3. Decision to stop
Illustrative decision path. No model accuracy figures or technical specifications are implied.

The context

An AI video-analytics solution was evaluated in a small group of retail locations. The pilot needed to establish whether the technology offered useful value in the operating environment.

The question to answer

The relevant question was whether the use cases, accuracy, and infrastructure conditions supported a worthwhile rollout. A working pilot alone would not settle that question.

My contribution

I defined the use cases, aligned stakeholders and the vendor, coordinated the proof of concept, and assessed the results.

The assessment

The pilot provided evidence about the solution’s value under the existing accuracy and infrastructure constraints. The assessment brought those practical limitations into the rollout decision.

The decision

The pilot was completed and concluded. No further rollout was planned because the solution did not offer sufficient value under the conditions assessed.

This was a decision outcome. There is no quantified savings or avoided-cost claim attached to it.

About this case

Generalised from the pilot summary available as of 20 August 2026. The account is intentionally concise because the source records the role, constraints, and conclusion without detailed evaluation data.

A related idea

Diagnosing a workflow before automating it.

Another piece of the work

From days to an hour in creative production.

View all cases