The AI consulting market is a mess. I can say that because I'm in it, and I spend a fair amount of time explaining to potential clients why they should be sceptical of the industry I belong to.
Lots of people woke up one morning, added "AI" to their LinkedIn headline, and started charging £2,000 a day for advice they got from ChatGPT the night before. I'm not exaggerating. I've seen the slide decks. Some of them are literally just reformatted AI outputs with a consultancy logo slapped on top.
So if you're thinking about bringing in an AI consultant, here's what you should expect, and what should make you run.
What good AI consulting looks like
A good AI consultant does four things.
They listen before they prescribe. If someone walks in with a solution on day one, they're selling you a hammer and calling everything a nail. The first step should always be understanding your business, your processes, your data, and your people. That takes time. Not months. But more than a one-hour discovery call.
I spend the first week of every engagement just talking to people and looking at systems. No recommendations. No slides. Just understanding. Because the right AI solution for a 30-person fintech is completely different from the right AI solution for a 200-person e-commerce company, even if they both say they want "AI agents."
They're honest about what AI can't do. If a consultant tells you AI can solve all your problems, they're lying. AI is genuinely powerful for certain types of work. It's useless for others. A good consultant will tell you "this is a great AI use case" and also "this one, don't bother, you need a better database first."
Last month I told a potential client that three of their five proposed AI projects weren't AI problems at all. They were data quality problems. Fix the data first, then we can talk about AI. They appreciated the honesty. We're working together now, starting with the data.
They build, not just advise. This is my biggest gripe with the traditional consulting model. They assess, they recommend, they leave. And you're stuck with a PDF and no way to execute.
Good AI consulting includes building. Prototypes at minimum. Working systems ideally. If the consultant can't build the thing they're recommending, question why they're recommending it.
They transfer knowledge. The goal should be making your team capable, not creating a dependency. If the consultant leaves and your team can't maintain or extend what was built, the engagement failed. Every piece of work I do includes documentation, a handoff session, and usually a month of support after I step away.
The red flags
I've collected these from conversations with companies who've had bad experiences. Some of them called me to clean up the mess.
"We'll build you a custom AI platform." Run. You don't need a platform. You need a solution to a specific problem. Anyone selling you a platform is selling you months of work that probably ends with something you could've built in two weeks using existing tools.
No technical depth. If the consultant can't explain how the AI actually works, at least at a conceptual level, they're a middleman. They're going to subcontract the technical work and charge you a margin. Ask them to walk you through how they'd build a specific solution. If they deflect to "our team handles the implementation," find out who "our team" is.
Buzzword density above 3 per sentence. If someone tells you they'll "deploy a multi-agent orchestration framework to operationalise your enterprise AI strategy," they're not speaking English. They're performing. Good consultants explain complex things simply because they actually understand them.
No reference clients. Everyone was new once, fine. But if someone's charging premium rates and can't point to a single company where they've done similar work, that's a concern. Ask for references. Call them. Ask specific questions: what did they actually deliver? Did it work? Would you hire them again?
Fixed solutions before understanding the problem. I had a company tell me their previous consultant recommended a vector database and a RAG pipeline in the first meeting. Before asking a single question about their data. That's not consulting. That's a sales pitch for their favourite technology stack.
What the engagement should look like
Here's roughly what to expect from a well-run AI consulting engagement, based on how I structure things.
Week 1-2: Assessment. The consultant talks to your team, maps your processes, assesses your data. This might be a standalone piece of work (like our AI readiness assessment) or the first phase of a longer engagement. Either way, this is where the real value starts. Not before.
Week 3-4: Prototype. Build something small. A proof of concept for the highest-priority use case. Not polished. Functional. Something the team can touch and react to.
Month 2-3: Build and iterate. Take the prototype, refine it based on feedback, integrate it into actual workflows, test with real data, measure the results.
Month 3+: Handoff and support. Documentation, knowledge transfer, training. The consultant should be actively working to make themselves unnecessary.
Total engagement length varies. A focused project might be 6-8 weeks. A broader AI transformation with multiple use cases might be 3-6 months. Anything over 6 months and you should be asking why. Either the scope is too big (break it up) or the consultant is milking it.
How much it should cost
Transparency on pricing, because nobody else seems to offer it.
A standalone assessment: £3k-£10k depending on company size and complexity.
A prototype build: £5k-£15k depending on the use case.
A full engagement (assess, build, deploy, hand over): £15k-£40k over 2-4 months.
If someone's quoting you £200k+ for an AI engagement at a company under 200 people, they're either massively over-scoping or massively overcharging. Probably both.
The question to ask yourself
Before you hire anyone, ask this: "Do we have a specific problem we want AI to solve, or do we just feel like we should be doing something with AI?"
Both are valid starting points. But they lead to different engagements. The first needs a builder. The second needs an assessor. Make sure you're hiring the right one.
If you want to talk about what this might look like for your company specifically, have a look at our services and get in touch. No pitch deck. No 47-slide proposal. Just a conversation about whether it makes sense.