Before you spend on AI, find out where it actually pays back.
Two to three weeks, independent, and deliberately willing to tell you that the answer is “not yet” or “not this”. You get a costed shortlist, a list of things to stop considering, and the reasoning behind both.
The board asked for an AI strategy. The pilots are not the answer.
Most organisations we meet already have three or four AI experiments in flight, each championed by a different team, none with an agreed measure of success. The question is not whether AI can do something impressive. It is which of the things in front of you is worth doing, in what order, at what cost, and which should be killed before it consumes another quarter.
Use cases chosen by enthusiasm
The loudest idea gets the budget. Nobody has compared it against the boring one that would save four hundred hours a year in a back office.
No honest read on the data
The proposal assumes clean, accessible, permissioned data. The reality is four systems that disagree and one spreadsheet that is authoritative for reasons lost to history.
Cost modelled once, at demo scale
Per-token economics look trivial at ten users and alarming at ten thousand. Nobody has drawn the curve, so nobody knows where it bends.
Five dimensions, scored, with the evidence attached.
Data readiness
What data exists, who owns it, how accessible it is, how good it is, and what it would cost to make it usable. Scored per candidate use case rather than as one blanket verdict.
Process readiness
Whether the process is stable enough to automate. An inconsistent, undocumented process does not become consistent because a model is pointed at it; it becomes inconsistent faster.
Process auditTechnical readiness
Integration surface, identity, environments, deployment, observability. Whether anything built could actually reach production in your estate, or would sit in a sandbox for a year.
Risk & governance readiness
Regulatory exposure, the EU AI Act, sector rules, data residency, the audit questions you will be asked, and who signs off when a model is wrong.
Standards & ISOPeople readiness
Who will run this after go-live, what they can already do, what they cannot, and whether the operating model has room for a system that needs supervision.
Value model
For each candidate: the hours or errors removed, the cost to build, the cost to run at real volume, and the payback period. Arithmetic, not adjectives.
Two to three weeks, mostly other people's time, not much of yours.
Frame
One session with the sponsor to agree what a good outcome would be and what constraints are non-negotiable. We also ask what has already been tried and what happened to it.
Gather
Six to twelve interviews across the teams who do the work, plus a look at the systems and the data. We sit with people doing the actual task rather than reading a process document about it.
Model
Score each candidate on the five dimensions, cost it at real volume, and sanity-check the value assumptions with the people who would feel the difference.
Report
A written report and a session with your leadership: the shortlist, the sequence, the costs, the risks and the candidates to drop, with the reasoning for each.
A document you can take to a board, not a slide with a rocket on it.
- A scored shortlist of use cases, in the order we would do them, with the reasoning visible.
- An explicit “do not do this yet” list — usually the most valuable page in the report, and the one that saves the most money.
- A build and run cost model at your real volumes, with the point where the economics change marked on it.
- The data and integration work that has to happen first, sized, so it can be planned rather than discovered.
- A governance checklist mapped to the obligations you are actually subject to.
- A one-page summary written for a non-technical board.
Questions we get asked
How is this different from the free consultation?▼
The free hour is a conversation: you describe the problem and leave with an honest opinion and a written summary. The assessment is a piece of work with interviews, evidence, a cost model and a report. Most people start with the free hour and decide from there whether the fuller assessment is worth it.
Will you recommend building something you would then be paid to build?▼
We will tell you what we think you should do, including when that is nothing, and including when it is work we do not do. If you would rather remove the question entirely, we are happy to be excluded from bidding on whatever the assessment recommends — say so at the start and we will put it in writing.
How much of our team's time does it take?▼
Roughly six to twelve interviews of forty-five minutes each, one kick-off and one closing session. We work around the people doing the job rather than asking them to prepare material for us.
What if the honest answer is that we are not ready?▼
Then that is the report, with the specific things that would have to change first and what each would cost. That outcome is common and it is not a failure — it is considerably cheaper to find out in three weeks than in three quarters.
Do you assess tools we have already bought?▼
Yes. If a platform or a copilot licence is already in place, we assess what it is genuinely delivering against what it costs, and whether the gap is the tool, the data or the way it was rolled out.
One free hour. No pitch, no obligation.
Bring the problem you are stuck on — an integration that keeps breaking, a cloud bill nobody can explain, an AI project that is all demo and no product, or a platform you are about to invest in. You will leave with a straight answer and a written summary, whether or not you ever work with us.
- Architecture and integration review
- AI feasibility — what will actually work, and what will not
- AWS, Azure and Google Cloud cost and design
- Business process audit and ISO readiness
- Technical due diligence before you invest or acquire