The short answer
Expect the early stages to be fixed and stated on the page: a half-day executive workshop, a two-week readiness assessment, a four-to-six-week strategy. Expect the build to be quoted once a design exists, and expect the quote to hold. Expect running costs to be monthly and to include the model usage, the monitoring and the person who answers when something drifts. A consultancy that cannot describe its pricing in those three shapes has not done this often.
This site publishes its own structure on the Packages page, with the figures marked as they are decided. We publish it because a buyer who understands the shape of the cost makes a better decision, and because the alternative, “contact us for pricing”, wastes everyone’s week.
What you are actually paying for
Consulting time is the visible line, but it is not the whole cost, and a proposal that shows only consulting time is hiding the rest. The full picture has four parts. The consultant’s time: workshops, assessment, design, build, testing, documentation. The platform: the licences or usage fees for the models and tools the design chooses, which you pay to the vendor, not the consultant. The internal time: your people in workshops, answering questions, testing and approving. And the running cost: monitoring, model usage as volume grows, and changes when your business changes.
Ask for all four in any proposal. A low consulting fee attached to a platform choice that costs you every month for years is not a low price.
What drives the cost up
- Data that is not ready. Paper, scattered spreadsheets, systems with no export. Every hour of tidying is an hour of consulting or an hour of your staff.
- Integration. Reading from one system and writing to another is where the engineering is. Two systems is a build; five is a programme.
- Consequence. Anything that acts on customers, money or records needs approval rules, testing and audit, and those are designed and built, not switched on.
- Regulation. Health, finance and any sector with a code adds governance work that has to be done properly or not at all.
- Scope creep by demo. Every impressive thing the consultant shows you becomes a requirement unless somebody holds the line. Hold the line; the first project should be small.
What drives it down
Using the platform you already pay for. Most businesses on Microsoft 365 or an equivalent own more AI capability than they use, and configuring it properly is cheaper than building beside it. Starting with an internal, low-consequence project. Having a named owner on your side who can make decisions in the meeting rather than after it. And, unfashionably, having decided what you want before the first conversation: a business that arrives with “we want to stop re-keying supplier invoices” spends less than one that arrives with “we want to do AI”.
Pricing patterns that should make you cautious
- A fixed price for a build before any design exists. Either the price is padded to cover the unknown or the scope will shrink to fit the price. Neither is in your interest.
- A percentage of savings. It sounds aligned; in practice the savings are unmeasurable and the argument about them lasts longer than the project.
- A platform fee bundled with the consulting so you cannot see which is which, or a consultant who is paid by a vendor to recommend that vendor.
- “Contact us for pricing” on everything. Some things are genuinely quoted; a workshop is not one of them.
- A long retainer proposed before anything has been built. Running costs make sense once there is something to run.
How to compare two proposals
Put them side by side on the four parts: consulting, platform, your time, running. Then ask each consultant the same two questions. What happens if the design shows the project should not proceed? (The right answer is that you pay for the design and keep it, and they tell you plainly.) And who owns the architecture documents at the end? (The right answer is you, in a form another consultant could pick up.) The proposal that answers both clearly is usually the cheaper one over three years, whatever the first-page number says.
What a first engagement looks like in elapsed time
Because much of the cost is time, it helps to know how long each stage takes when it is run well. An executive workshop is half a day, with a short call and a questionnaire before it. A readiness assessment is about two weeks, most of it the consultant looking at systems and permissions rather than sitting in your meetings. A strategy engagement runs four to six weeks, with two or three workshops and a written roadmap at the end. A design stage for one use case is two to four weeks. A first build is six to twelve weeks, depending on how many systems it touches, followed by a month of running it with a person checking the output.
Two things follow. First, a consultant who proposes to build before any of the earlier stages exist is skipping the cheap weeks to get to the expensive ones. Second, your own time is real: budget a few hours a week from the owner of the process throughout, and more during the workshop weeks, because an engagement that runs without you produces a design for a business that does not exist.
Where to go from here
The Packages page on this site shows our four packages and the shape of each price. The AI Strategy service is the one that turns “we want to do AI” into a ranked list with a business case for the first item, which is the document that makes every later cost smaller. And the AI Opportunity Score below is free, takes five minutes, and tells you which package you are actually in the market for.
Published 12 September 2026 · Be AI