How to become an AI consultant: the honest version

Start with what the salary number really says

IT Jobs Watch, which tracks UK IT job adverts, puts the median advertised salary for an AI consultant at £82,500 for the six months to 9 September 2026, up 6.45% on the same period a year earlier. That's a good living and a real trend.

The number underneath it matters more. That median comes from 120 postings across six months. For a settled job title you would expect thousands. The honest read is that AI consultant is still an emerging label rather than an established grade, and the figure reflects scarcity as much as seniority.

You will also see freelance day rates quoted anywhere from a few hundred pounds to well over a thousand. Those rate cards are self-reported by people selling consulting, so treat them as marketing rather than data. What you can charge depends almost entirely on whether you can point at work that worked.

What the job is, most days

The mental image most people have is someone building models. That's an AI engineer or a data scientist, and it's a different job with a different skill set.

Consulting is mostly translation. You sit with a team, watch how a piece of work actually gets done, and work out which part of it a model can take and which part it can't. Then you make the case to someone with a budget, in their language, with numbers they'll accept.

Concretely, a week looks more like this: interviewing people about their workflow, mapping where the time actually goes, building a rough prototype to test whether the idea survives contact with reality, and writing it up so a sceptical finance director will sign. The part that involves AI is maybe a quarter of it.

The line between this and a training engagement is worth understanding before you pick a lane. We wrote it up in AI consultant vs AI training provider, and what a Head of AI does covers the in-house version of the same job.

The two routes in, and the one people undervalue

Almost everyone arrives from one of two directions.

• From technology. You already build software or work with data, and you add the business-facing layer: scoping, stakeholder management, commercial writing, knowing when to stop building. The gap here is commercial, not technical, and people consistently underestimate how large it is.
• From a domain. You're an accountant, a lawyer, a recruiter, an operations manager or a marketer with ten years in the work, and you add genuine AI fluency. The gap is technical, and it's smaller than it looks from the outside.

The domain route is the undervalued one. Organisations are not short of people who can explain how a language model works. They're short of people who understand how a claims process, an audit file or a month-end close actually runs, and who can also see where a model fits into it. If you already have the domain, you're closer than you think.

What I'd tell someone starting now

For context on where I'm arguing from: I've worked on AI since 2008, when I wrote a chatbot dissertation for an MSc at Manchester, which at the time was an unfashionable thing to do. Then I spent years in finance at Trafigura, BNP Paribas and Hitachi Capital, started Apexure in 2015 and Fautons in 2024.

The useful part of that for consulting was never the AI credential. It was the years of watching how businesses actually run, and how decisions actually get made. So the advice I'd give my younger self isn't to chase the AI part first.

• Pick one domain and go deep, or use the one you already have. Generalist AI advice is a crowded and low-margin place to stand.
• Get properly fluent in two tools rather than shallow across ten. Fluency means knowing the failure modes, not the feature list.
• Build three things end to end and write up what happened, including what didn't work. That's your portfolio, and it does more for you than any certificate.
• Learn to write a proposal and scope a piece of work. Most people who struggle at consulting struggle here, not at the technology.
• Find the first client through people who already trust you. Almost nobody's first engagement comes from a cold search.
• Get comfortable telling someone a project shouldn't happen. Talking a client out of a bad idea once will do more for your reputation than three successful builds.

About those certifications

Search for this and you'll get a page of certification programmes, several describing themselves as accredited. It's worth understanding what that word is doing.

Accreditation means a recognised external body has assessed the programme against a standard. Plenty of AI certificates use it loosely, sometimes meaning no more than paid membership of a trade association. There is no chartered body for AI consultants in the UK the way there is for accountancy, law or engineering, so nobody is in a position to award the equivalent of a practising certificate.

That doesn't make structured learning useless. A good programme gives you a curriculum, deadlines, feedback and something to show at the end, and those are genuinely worth paying for. What it won't do is make a client hire you, because clients buy evidence of work rather than evidence of study.

We run an AI for Professionals certification ourselves and we deliberately don't call it accredited, because it isn't. It's built around a portfolio project for the reason above, and it's on a waitlist while the first cohort is put together. Is an AI certificate worth it is the longer version of this argument, and if your employer might fund the learning, can I expense AI training on my learning budget covers how to ask.

The part nobody puts in the job description

Consulting is a sales job with a delivery component attached. You can be the strongest practitioner in the room and earn nothing, because nobody knows you exist. That isn't a cynical observation, it's the thing that decides whether this works as a career.

The practitioner skills that hold up are the ones in AI skills every professional needs: judgement, workflow design and verification. Those make you good at the work. Getting paid for the work is a separate discipline, and the sooner you start practising it, the shorter the unpaid stretch at the beginning.