AI Automation Assessment

Who it is for

The AI Automation Assessment fits when a team is spending hours each week on a task that looks automatable and you want a payback figure before anyone builds a demonstration.

• A team spends hours each week on triage, extraction, drafting or routing that follows a pattern.
• The board has asked what the business is doing about AI and the honest answer so far is not much.
• An earlier pilot produced a convincing demonstration that never reached production.
• You want to know whether the data behind a process can be read reliably before spending anything on a model.

If you already know the workflow and want it built, the AI automation pillar carries the published band for the build itself. If the question is broader than one workflow, the Technology Roadmap sequences the whole estate first.

What you get

A written pack about a week after the workshop, and an hour of playback to go through it.

• The process as it is actually performed, mapped with the people who do it.
• Findings on the systems the task depends on: whether their data can be read reliably, whether an answer can be written back, and what rate limits or approval steps sit in the way.
• An agreed definition of good enough, which the build then scores against real historical cases from your business.
• A shortlist of candidate workflows ranked by payback period.
• A fixed quote for the first automation, and what it would take to run it.

The definition is the part most assessments skip. Without one there is no way to tell a system that works from one that reads well, and scoring it against your real cases is the first step of the build. The pack is yours, and it is written so that another team could build from it.

How it runs

1. A 30-minute call. We confirm which process the assessment examines and who needs to be in the workshop. Nothing is invoiced before this call.
2. The workshop. Two hours with the people who do the task, on the work as they actually do it.
3. Data and systems check. Whether the inputs the task depends on are reachable and clean enough to automate against, and where an answer has to land.
4. The definition of good enough. Agreed with you in plain terms, so the build can be scored against the outcome a person produced on the same input.
5. The pack and playback, about a week in. The ranked shortlist, the fixed quote and an hour on the reasoning.

An AI position for the board

Boards increasingly ask for a written position on how a business uses AI: which tasks are automated, what a person still approves, where data goes and how a wrong answer is caught. That document can be added to the assessment and is quoted separately. It draws on the same findings, so it describes what the business actually does rather than restating a generic policy.

What happens after

Where the shortlist has a workflow worth building, the fixed quote is for that build and the assessment fee is credited in full against it if it proceeds within 90 days. Where the answer is that a workflow should be left alone, usually because the data behind it cannot be read reliably or the volume does not justify the build, that is written into the pack. It is far cheaper to learn here than during a build.