£2,950 fixed · five working days
Copilot doesn't fix your model. It inherits it.
Switch AI on over a model that disagrees with itself and you don't get a reporting problem, you get a broadcasting problem: the same errors, delivered faster and with more confidence, to people who have no way of checking. This tells you whether your estate is ready, and what it costs to run when it is.
A scored readiness report covering naming, descriptions, synonyms, relationships and measure trust, model by model.
An evaluation harness: thirty real business questions with verified expected answers, run before and after any changes, so readiness becomes a pass rate rather than an opinion.
A ranked remediation list, so you know what to fix first and what can wait.
Copilot capacity and licence cost modelling, so the Fabric bill is a decision rather than a surprise.
Tenant settings and access recommendations, including who should not be given Copilot yet and why.
A 90-minute readout for whoever is sponsoring the AI spend.
Questions first, model second
We write the thirty questions with your team: the things people actually ask, in the words they actually use, with the right answer worked out from source. That list is the standard everything else gets measured against.
Score the models, run the harness
Every model gets checked against the readiness standard, then the thirty questions get run so you can see exactly where the answers go wrong and why.
The readout
One report, the pass rate, the ranked fix list and the capacity numbers. Enough for you to decide whether to switch Copilot on, delay it, or restrict who gets it.
- Up to three semantic models, £650 for each one after that
- One Power BI tenant
- Scored readiness report against a published checklist
- Fee credited in full against remediation booked within 90 days
- A free 20-minute call, no pitch
- We work out which engagement fits, or whether none of them do
- If a cheaper piece of work solves it, I'll tell you
A wrong number nobody spots is worse than no number.
Reporting has always had a trust problem, and it has always been survivable, because a person reads a report and thinks "that looks off". A conversational answer removes that step. Nobody sees the model, the filters or the measure it chose, so nothing looks off. It just sounds right.
That's why this assessment exists, and why it is built around a tested question set rather than a checklist of features. The question is never "does Copilot work", it's "does it agree with your accounts".
50% to book, 50% on delivery, 14 days to pay.
Invoice or card, whichever suits your finance team. All prices exclude VAT where applicable.
The deposit holds your date
Booking blocks out my diary, so the first half is non-refundable inside 14 days of the start date. Move it with more notice than that and nothing is lost.
Work pauses on an overdue invoice
Not a threat, just how it works, and saying it now saves an awkward conversation later.
Scope changes get quoted, not absorbed
Every fixed price names what it covers. Anything outside that gets priced before it starts rather than appearing on a final invoice.
Access delays move the date, not the price
Waiting on permissions is the most common reason a turnaround slips. It shifts the delivery date and never the fee.
Frequently asked questions
What is an AI-Ready Power BI Assessment?
It's a fixed-price £2,950 assessment of whether your semantic models will give Copilot correct answers. Copilot does not fix a model, it inherits one: ambiguous field names, missing descriptions, no synonyms and measures nobody trusts all turn into confident wrong answers, faster than a human could produce them. The assessment scores what is there, lists what to change, and tells you what the capacity will actually cost.
What is the evaluation harness?
Thirty real business questions, written with your team, each with a verified expected answer worked out from source. Those get run against your model before any changes and again afterwards, so AI readiness stops being a matter of opinion and becomes a pass rate you can see move. It is also re-runnable, which makes it the natural thing to re-test after any significant model change.
Do I need Fabric capacity before you can do this?
No, and finding out what you would need is part of the point. Copilot in Power BI needs an F2 capacity as a floor, but sustained real use pushes much higher, and the gap between those two numbers is where budgets get embarrassed. You get capacity and licence cost modelling as a named deliverable, so the finance conversation happens before the bill does.
Is this just a Copilot demo?
No. Plenty of Copilot workshops are a day of slides and a demo on sample data. The deliverables here are specific to your estate: your models scored, your tenant settings reviewed, your questions tested, your capacity costed, and a written list of who should not be given Copilot access yet and why.
What if the answer is that we are not ready?
Then that is the finding, and it is worth more than a green light would be. Switching Copilot on over a model that disagrees with itself does not create a reporting problem, it broadcasts one. The report tells you what to fix first and what it would take, and the fee credits against that remediation work if you book it within 90 days.
Stop arguing about the numbers. Start using them.
Book a fixed-price Power BI Health Check and find out how trusted, usable and audit-ready your reporting really is, and exactly what to do next.