How much does an AI consultancy cost?

"How much will it cost?" is the first question we get on almost every AI enquiry, and it's the one that gets the least useful answer online. Plenty of agencies publish confident-looking price tables. The truth is that a two-week readiness review for one team and a six-month rollout across a whole business are so far apart that a single number tells you nothing. What you can pin down is how a good consultancy prices, what shapes the bill, and how to make sure you're not overpaying.

Two ways consultancies price AI work

Most UK AI consultancies price in one of two ways: day rates for time-and-materials engagements, or fixed-price stages for well-scoped work. Both are legitimate. Fixed-price is usually better for you when the scope is clear (a readiness assessment, a Copilot pilot, an agent build). Day rates make sense when the work is genuinely exploratory and locking down a price would just build in a large contingency.

Wherever you land, the number that matters is the one that reflects your scope - not a headline range copied from somebody else's project.

Typical project shapes

Rather than quote ranges that won't apply to you, here's the shape of the work we most often see. Any consultancy worth talking to will size each stage against your actual environment:

• An AI readiness assessment. Two to three weeks. Licensing, security, data hygiene and use case review with a written recommendation.
• A Copilot pilot for one team. Six to ten weeks. Deployment, adoption, prompt libraries and measurement.
• A Copilot Studio agent build. Sized around the process complexity, integrations and evaluation work needed.
• A wider AI rollout. Three to six months. Multiple teams, governance, adoption programme and reporting.
• A licensing and estate review. A short focused piece that often pays for itself in licence savings alone.
• Ongoing AI care. Monthly, scaled to how many enhancements and new agents you want in the queue.

Don't forget the licences

Consultancy fees are one line on the bill. AI platform licences are the other, and they behave very differently:

• Microsoft 365 Copilot - per user per month, on top of a qualifying Microsoft 365 plan.
• Copilot Studio - consumption-based (message packs), scaling with usage.
• Azure OpenAI - pay-as-you-go by tokens; usage varies enormously by use case.
• Microsoft Fabric - capacity-based, sized to the workload.
• Microsoft Purview - varies by add-on; sensitivity labelling is usually included in existing E3 or E5 licences.

Right-sizing the licence model is one of the highest-value things a good consultancy does for you. It's very common to see clients over-provisioned because someone bought Copilot for everyone when only half the team would use it, or bought Fabric capacity several sizes too big.

What drives the bill up

• Data readiness. If your SharePoint and Teams estate is a mess, or nobody knows where the important documents live, discovery and clean-up take longer.
• Oversharing risk. Copilot can see everything a user can. Fixing permissions properly before rollout is real work, and it's not optional.
• Custom agents. A bespoke Copilot Studio agent with integrations, evaluation and guardrails is a proper build, not a weekend prototype.
• Retrieval-augmented generation. Grounding a generative AI model on your own content adds indexing, chunking and evaluation work.
• Change management. Adoption, training and running a governance model cost real money but pay back the most.

What drives it down

• Clear success measures agreed up front - "what does 'done' look like?"
• A small, motivated pilot group instead of a company-wide big-bang.
• Using what Microsoft already gives you (Copilot in the apps, Power Automate flows) before building something bespoke.
• A fixed-price stage-by-stage engagement so you can stop, redirect or continue with your eyes open.
• Being honest about which processes actually hurt - AI on the wrong process is expensive and unimpressive.

How we price

At Axon we scope in fixed stages - discovery, readiness, pilot, rollout, ongoing care - so you're never signing a blank cheque. Every number we quote is tied to your actual scope, not a headline range designed to catch attention. Tell us what you're trying to do and we'll come back with a proper price for that piece of work, along with what we would and wouldn't do next.

Request a tailored quote

The only honest way to give you a number is to look at your actual scope. Answer a few quick questions and we'll come back within one working day with a tailored view of what makes sense, what it would cost, and what we'd suggest doing first. No demo, no chase.

Where to go next

If you want to see what an engagement would look like for your business, our AI consultancy page walks through how we work. For the Microsoft-specific view, see our Microsoft AI consultancy page, or our Copilot consultancy page if Copilot is where you're starting.