How we work · Prove
Draw the line with evidence.
Before anything is built into your business, the same real work runs three ways, side by side. The results show what should stay human, what should get AI support and what can run on its own.
Mid-market engagements. On the SME side, Prove is a simpler side-by-side check.
01Why it matters
Evidence replaces opinion.
Most decisions about what to automate rest on opinion: a vendor demo, a pilot on clean data, a strong view in the leadership team. Evidence from your own work replaces the guesswork, and the people affected see the same numbers you do.
02The parallel test
The same work, three ways.
The same real work, done three ways: people alone, people with AI, and AI alone. We compare time, errors, exceptions and how often people overrule the AI, on samples large enough to trust.
- 01
People, unassisted
The work as it runs today. This is the baseline everything else is measured against.
- 02
People with an AI co-pilot
The same work, with a co-pilot supporting the team. The AI drafts, checks and suggests. Your people accept, change or reject what it produces.
- 03
A standalone AI agent
A working prototype tests what can run on its own. The agent does the work end to end. Every output is reviewed before it goes anywhere, so nothing reaches a client or a system without a person's approval.
What we measure
- 01Hours per case, including human review and correction time
- 02Error rate on comparable cases
- 03Exceptions
- 04How often people overrule the AI
- 05Operating cost
- 06The acceptance thresholds agreed in Discover
03What you get
A decision for every step, written down.
For each step of the workflow, a decision backed by evidence: it stays human, it gets AI support, or it runs on its own with review. Those decisions become written decision rights, showing where AI acts, where it advises and where a person decides.
The decisions recorded: where AI acts, where review is required and who handles exceptions.
How your team takes part. They take part in the test and see the results as they come in. The decision about their work is made with them, on numbers they recognise, rather than handed down.
When it runs. During Prove, on your real work, after Discover has set the baseline. We agree the timeline with you before you commit. How we work
Start with one function and one metric.
A 45-minute conversation to agree where the programme would start and what it would move.