What does an AI consultant actually do?

Dovi Levin Founder, Zmanify · 16 July 2026

An AI consultant works out where AI genuinely fits your business, then makes it stick: finding the tasks that eat your team's hours, matching them to the right tools, and training people on their own work until the time savings are real. The good ones also tell you where AI doesn't fit. And sometimes the honest answer is that you don't need one at all.

What does the work look like day to day?

Strip away the job title and the work is four things:

• Teaching. Giving a team a working understanding of AI: what it actually does, where it's useful, and where it's just hype. In plain English, built around their work, not a generic webinar.
• Auditing. Walking through how the business runs day to day, finding the bottlenecks and manual work, and pinpointing where AI saves hours. We've written up exactly what happens in an AI audit if you want the detail.
• Hands-on training. Taking the team's real tasks and running them through AI live, then having people do it themselves. This step is where change happens or doesn't.
• Building small tools. Sometimes a recurring task justifies a custom AI assistant: something that drafts the weekly report or triages the shared inbox. Small, boring, and worth hours a week.

Notice what's missing: writing code for months, replacing your systems, or producing a 60-page strategy. If a proposal is mostly those, you're talking to a software agency or a management consultancy wearing an AI badge.

A typical engagement runs in that order, too. First a session that gets everyone on the same page, sceptics included. Then the audit, which turns "AI is interesting" into a ranked list of your specific opportunities. Then the hands-on part, where your team watches their own work happen faster and learns to do it themselves. The whole arc can fit inside a few weeks; the point is a team that no longer needs the consultant, not a retainer that never ends.

How is that different from a developer or an agency?

Developers build software. Agencies hand you a deck. An AI consultant works at the level of your operations: your inbox, your reports, your month-end, your client work. The output is measured in hours returned to your team, not in features shipped or slides delivered.

That difference matters when you buy. A developer needs a specification. An agency needs a brief. An AI consultant needs to watch how your business actually runs, which is why any engagement that skips that step is guesswork with an invoice.

What does an AI consultant cost?

In the UK in 2026, freelance AI consultants typically charge £400 to £800 a day, boutique specialists £1,200 to £2,500, and the Big Four upwards of £1,500. A first engagement for a small or mid-sized business, usually an audit, commonly starts at around £750. We've broken down the full cost picture separately, including the add-on costs to ask about before you sign anything.

The more useful question is the return. One person saving three hours a week is roughly 150 hours a year. Price the engagement against the hours it returns, not against the software budget.

How do you choose a good one?

Four tests, all askable in the first call:

1. Do they ask about your business before mentioning tools? Anyone who leads with a product name is selling that product.
2. Do they talk in hours and pounds? "Improved productivity" is not a deliverable. "Four hours a week off your month-end" is.
3. Will they say where AI doesn't fit? Ask directly: "What shouldn't we use AI for?" A good consultant has a fast, specific answer. A salesperson changes the subject.
4. Do they train your people or create dependence? The goal is a team that works without the consultant. Be wary of anyone whose plan makes them permanent.

When you don't need an AI consultant

Plenty of businesses don't, and it's worth saying plainly:

• You're a one-person business with simple workflows. A £20-a-month subscription and a few focused weekends will get you most of the way. Start there.
• Your team is already experimenting happily. If people are sharing prompts and workflows on their own, pour fuel on that instead of hiring someone to restart it.
• Your processes are broken. AI applied to a chaotic process makes the chaos faster. Fix the process first; that might need an operations person, not an AI one.
• You'd be spending money you can't spare. If the fee would genuinely hurt, wait. The tools get cheaper and more capable every quarter, and the free tier of ChatGPT or Claude is enough to learn on.

A consultant earns their fee when a business is past the point where any one person can see where all the time goes, or when a year of "we should really look at AI" has produced licences but no change. If that second one stings, we wrote about why teams with ChatGPT licences see nothing change.

Where do you start?

Before you speak to any consultant, us included, get a picture of where your hours actually go. Our two-minute quiz does exactly that and emails you a personalised report on where AI is most likely to save you time.

Then, if it looks worth a conversation, book a call. If you're one of the businesses that doesn't need a consultant, we'll tell you that too. It's quicker for everyone.