Straight answer: DIY AI tools work brilliantly for single, well-defined tasks like writing first drafts or transcribing calls, but they fall apart the moment you need several tools talking to each other or a strategy behind them. Hiring an AI expert costs more upfront but usually pays for itself within a few months once you count the hours you’d otherwise burn on trial and error. The right answer for most small businesses is a mix of both, not a straight choice between them.
What DIY AI costs you (it’s not the subscription)
Everyone quotes the ChatGPT Plus price, £20 a month, and stops there. That’s not the real cost. The real cost is the 30 to 60 hours most business owners spend working out which tool does what, why their prompts keep producing generic rubbish, and how to stitch three separate tools together so the output from one feeds cleanly into the next.
I had a client, a boutique PR agency owner in Manchester with four staff, who spent close to three months building her own content workflow using ChatGPT, Canva, and Zapier. She’s smart, she’s not lazy, and she still ended up with a system that broke every time OpenAI changed something small in the background. When she finally brought me in, we fixed the whole thing in nine days. Not because I’m cleverer than she is, but because I’d already made every mistake she was about to make, on someone else’s time and budget.
That’s the bit people skip when they compare “£20 a month” against “£2,000 for a consultant.” The £20 tool cost her roughly £4,000 in her own billable hours before she gave up and paid for help anyway. DIY isn’t free. It’s deferred cost, and it usually comes with interest.
What hiring an AI expert gets you
Here’s the uncomfortable bit most people avoid saying plainly: a decent AI consultant is not doing anything you couldn’t technically learn to do yourself. There’s no secret AI that only consultants get access to. What you’re paying for is someone who has already tested forty dead ends so you don’t have to walk into them yourself, and who can tell you honestly when a tool is wrong for your business before you’ve sunk six weeks into it.
A good expert will:
- Audit what you’re already doing manually and tell you which parts are worth automating first (usually the boring repetitive ones, not the flashy ones)
- Pick tools based on your actual team size and budget, not whatever’s trending on LinkedIn that week
- Set up the connections between tools so information moves without you copying and pasting
- Train your team so the system doesn’t collapse the day the consultant leaves
- Tell you, sometimes, that you don’t need AI for a particular problem at all
That last point matters more than people think. I’ve turned down work where a business owner wanted an AI chatbot when what they needed was a proper FAQ page and a human answering the phone for another hour a day. If you’re weighing this up for customer service specifically, it’s worth reading what outsourcing live chat really involves for small businesses before you assume AI is the fix.
The maths, with real numbers
DIY stack for a small business doing content, admin, and basic customer replies: ChatGPT Plus or Claude Pro (£20 a month), a transcription tool like Otter.ai or Fireflies (£10 to £20 a month), Zapier or Make for connecting things (£20 to £70 a month depending on volume). Call it £50 to £110 a month, plus your own time learning it, which is the variable nobody puts in the spreadsheet.
Hiring help ranges wildly depending on what you need. A single strategy session with an AI consultant might run £300 to £800. A short project to set up and automate one workflow typically lands between £1,500 and £5,000. Ongoing fractional support, where someone effectively acts as your part-time AI lead, usually sits between £1,000 and £4,000 a month depending on hours and scope. If those numbers feel abstract, I’ve broken down the actual ranges and what drives them in more detail on how much an AI consultant really costs.
Here’s the calculation that matters: if your own time is worth £40 an hour or more, and DIY implementation is realistically going to eat 40 to 80 hours before it works, you’re looking at £1,600 to £3,200 of your own time against a project fee that often lands in the same range but gets finished in a fortnight instead of a quarter. That’s not an argument for always hiring. It’s an argument for doing the sum before you decide.
A simple way to decide, step by step
- List the specific task, not the vague ambition. “I want to use AI” isn’t a task. “I want first drafts of client emails written in ten minutes instead of forty” is.
- Check if one tool solves it end to end. If yes, this is a DIY job. Learn the tool, budget a week, move on.
- Check if it needs two or more tools to talk to each other. This is where most DIY attempts fail. If the task involves a handover between systems, budget for help or budget serious learning time.
- Work out what your time is worth per hour, then multiply by a realistic 40 hours of trial and error. Compare that number honestly against a quote.
- Ask any expert you’re considering what they’d say no to. If they say yes to everything you propose, keep looking.
Step five is the one people skip, and it’s the one that saves you the most money. A useful checklist of what to interview for lives in what to check before hiring an AI consultancy for marketing, and a lot of it applies beyond marketing specifically.
Where DIY quietly fails, even for capable people
The pattern I see most often isn’t stupidity, it’s scope creep. Someone starts with one simple use case, it works, they get confident, and within a month they’re trying to automate their entire sales pipeline using free tiers and tutorials from people who’ve never run a business with a payroll to meet. The tools themselves don’t warn you when you’ve gone past what they’re built for. ChatGPT will happily help you design a customer database structure it has no business helping you design.
That’s the honest gap in most “just use ChatGPT” advice online. The tool doesn’t know your business, doesn’t know your compliance obligations if you’re handling customer data, and won’t tell you when you’ve built something fragile. A person who’s done this for other businesses will.
This is roughly the point where a lot of owners start looking at bringing in dedicated help rather than another tool, and it’s worth understanding what that looks like day to day. I wrote about this in what a fractional AI officer is and when a small business needs one, because “hiring an AI expert” doesn’t have to mean a full-time hire or a one-off consultant. There’s a middle option most people don’t know exists.
When DIY wins
I’m not building a case for always hiring, because that would be self-serving and wrong. DIY is the right call when:
- You’re testing a use case for the first time and don’t yet know if it’ll stick around long enough to justify a fee
- The task is contained to one tool with no handoffs to other systems
- You or someone on your team already has a bit of technical comfort and enjoys tinkering (this matters more than people admit)
- Your budget can’t stretch past £100 a month right now and the alternative is doing nothing
If any of those apply, spend a focused week, not a scattered three months, learning one tool before you touch a second one. Most failed DIY attempts I’ve seen weren’t failures of the tool, they were failures of trying to learn five things simultaneously.
The blended approach that works for most small businesses
What I recommend to almost every client now, and what I did for my own business during the rebuild years, is start DIY on the cheapest, lowest-risk task you have, learn what “good” looks like by doing it badly a few times yourself, and then bring in expert help specifically for the piece that connects systems together or touches customer-facing risk. That’s usually a two or three week engagement, not an ongoing retainer, unless your business has grown enough to need continuous support.
If you’re weighing up a firm rather than an individual, the questions don’t change much by geography, though I’ve written a location-specific version for anyone comparing options locally in what to know before hiring an AI consulting firm in Brisbane. The principles travel fine outside Australia.
For anyone still deciding between a fully DIY route and getting proper support, the most useful next step is usually a short paid consultation rather than either extreme. You can see what that looks like in practice through working with an AI consultant for a small business, which walks through what a first engagement typically covers before you commit to anything bigger.
Frequently asked questions
Can a small business really do AI implementation without any outside help?
Yes, for single, contained tasks like drafting content, summarising calls, or basic data cleanup. It gets much harder once you need two or more tools connected or you’re handling customer data across systems, which is where most DIY attempts stall out around week six or eight.
How much should a small business expect to spend on an AI expert?
A single strategy session typically costs £300 to £800, a defined project runs £1,500 to £5,000, and ongoing fractional support usually falls between £1,000 and £4,000 a month depending on scope and hours involved.
Is it cheaper long term to learn AI tools yourself instead of hiring someone?
Sometimes, but only if you count your own time at its real hourly value. Most owners underestimate the 40 to 80 hours DIY implementation takes and end up paying more in lost time than a project fee would have cost.
What’s the biggest mistake small businesses make when choosing between DIY and hiring?
Trying to learn several tools at once instead of mastering one, and hiring for the wrong stage of the problem, bringing in expensive strategic help for a task a £20 a month tool could have solved on its own.
Related reading: Expert Advice on Hiring Full Stack and AI Consultants for UK Small Business in 2026: The Honest Hiring Guide From 21 Years on the Front Line.
I go much deeper on this in the AI marketing guide.
