What does an AI consultancy actually do?
Strip away the hype and an AI consultancy does three things: finds where your business loses money to manual work, proves the value of fixing it, and builds the fix. Here's what that looks like in practice.
Ask ten firms what an AI consultancy does and you'll get ten answers, most of them involving the word "transformation". Here is the honest version, from one that builds systems for a living.
What does an AI consultancy actually do day to day?
1. Diagnosis. Most businesses don't know where their hours go. The consultancy's first job is to follow real work through the business: who touches an order, an invoice, an enquiry, and where it waits. The output should be specific: a map of the process and a priced list of where time and margin leak. If a consultancy can't tell you what a broken process costs you per month, they're guessing.
2. The commercial case. Not every automation is worth building. A good consultancy models the current cost, the delivery cost, and the expected return, then tells you which ideas fail the test. This is the step most firms skip, and it's the one that protects you. Weak ideas should die on paper, not in a six-month project.
3. The build. Working software wired into the tools you already use, not a strategy deck, not a proof-of-concept that never leaves a sandbox. The test is simple: after the engagement, is a system running in production that wasn't before, and can you point at what it does each week?
What you should get for your money
- A process map of how work actually moves through your business, not the org-chart version.
- Opportunities ranked by return, each with the hours it recovers and what those hours cost you today.
- A costed recommendation you can act on with the consultancy or without them.
- If you proceed to a build: software in production, documentation, and a named owner inside your business.
Which questions separate a real AI consultancy from theatre?
- "Can I see something you've built that's running today?" Case studies with screenshots and numbers, not logos and adjectives.
- "What would you not automate in my business?" A real consultancy has a list. Anything involving judgement, relationships or irreversible decisions usually keeps a human in charge.
- "Who owns what you build?" The right answer is you.
- "What happens after launch?" Automation needs monitoring and improvement. Ask what the monthly rhythm looks like.
When don't you need an AI consultancy?
If your problem is already solved well by off-the-shelf software (accounting, CRM, scheduling), buy the software. A consultancy earns its fee where your process is specific to you, crosses several tools, or where the off-the-shelf options force your business to work their way instead of yours. M22 wrote more on that decision in build vs buy.
How does M22 help with this?
M22 Consultancy runs exactly this diagnosis-case-build sequence, under the method Understand, Quantify, Build, Improve: the AI Audit maps real work and prices the leaks, ranks opportunities by expected return, and only then does M22 build anything, for a fixed fee agreed before work starts. The client owns the code, data and documentation, and a monthly retainer keeps M22 measuring adoption and improving the system once it is live, rather than stopping at advice. The AI Audit starts at £1,500, a fixed fee agreed before day one. Book a thirty-minute call, or see how to choose an AI consultancy.