The method
A single, repeatable way of deciding where artificial intelligence earns its place in a business — and where it does not.
01
How the work actually gets done.
Two weeks of observation and interviews across the processes that carry your cost and your risk: quoting, invoicing, intake, service, handover. We document the work as performed, not as described in the process manual.
02
Every candidate use-case, on the same four axes.
Value, effort, data readiness and risk. Each candidate is scored identically, so the shortlist is comparable rather than persuasive. Anything that fails the data-readiness axis is named as a prerequisite instead of a project.
03
Money in, money out, in writing.
Licence, integration, change and oversight cost against a defensible estimate of return. If a use-case cannot be costed honestly, it does not enter the roadmap.
04
Ownership before deployment.
A named owner, a human review step, an entry in your AI register and an EU AI Act risk classification for each initiative that proceeds. Governance is written at the start of the engagement, not retrofitted after the pilot.
The four scoring axes
Measurable effect on cost, cycle time, quality or capacity.
Integration, change management and oversight burden.
Whether the data required already exists and agrees with itself.
Regulatory classification, privacy exposure and failure cost.