The short version: Start with one skill you can already prove, one type of small business you understand, and one paid pilot project, not a website and a logo. Most people who fail at this spend three months "positioning" themselves and zero hours talking to a client. The money starts when you stop preparing and start doing a small, cheap, slightly embarrassing first job.
Worth reading next: AI Automation for Agencies: Deliver More Without Hiring.
Why this side business is worth building right now
I get asked about this most weeks, usually by someone in their 40s or 50s who's spent twenty years in marketing, operations, HR or finance and has watched their employer half-heartedly roll out an AI tool nobody trained anyone to use. That person already knows more about ChatGPT, Claude or a basic automation workflow than 90% of the small business owners in their town. That gap is the entire business.
Small businesses in the UK and US are not short of AI hype. They're short of someone who will sit with them for two hours, look at their actual invoicing process or their actual customer emails, and say "here's the three things worth doing, in this order." That's the job. It's not glamorous. It pays well because almost nobody is doing it at the small business level, they're all chasing enterprise contracts or building courses.
Step 1: pick a lane before you pick a logo
The biggest mistake I see is people calling themselves an "AI consultant" full stop, which means nothing to a plumber, a dental practice or a 12-person accountancy firm. Pick one of these to start:
- Customer service and email response automation for service businesses
- Content and marketing workflows for agencies or solo marketers
- Admin and scheduling automation for clinics, salons or trades
- Sales follow-up and CRM cleanup for small B2B firms
- Internal knowledge and document search for firms with 10 to 40 staff
Pick the one where you already have five years of context. If you spent a decade in recruitment, don't try to break into legal tech. Go straight to recruitment agencies with 5 to 30 staff, they'll trust you faster because you speak their language, and you'll close deals in weeks instead of months.
Step 2: build one proof piece, not a portfolio
You don't need ten case studies. You need one. Pick a real, small, unglamorous problem, ideally in a business you can access for free, your own, a friend's, a former employer, and fix it publicly.
Here's the one I still tell people about. In 2023 I sat with a friend who runs a three-person bookkeeping firm in Kent. Her team was spending roughly six hours a week manually writing client update emails, the same five templates, tweaked slightly each time. We built a simple system using ChatGPT with saved prompts and their existing client data pulled from a spreadsheet. Setup took four hours. It cut that six hours a week down to under ninety minutes. That's it. Not a transformation, not "AI-powered growth", just ninety minutes back for a business owner who valued her time at roughly 40 pounds an hour. That's 3,800 pounds a year of recovered time from four hours of my work.
Write that up as a one-page case study: the problem, what you did, the before and after number, nothing else. That single page will do more for your credibility than a full website with a "services" dropdown menu.
Step 3: price it so it's easy to say yes
Your first three clients should not be your normal rate. This isn't about undervaluing yourself forever, it's about getting real feedback and real referrals fast. A common structure that works:
- Pilot project: fixed fee, 300 to 800 pounds, one specific process fixed in one to two weeks
- Follow-on audit or setup: 800 to 2,000 pounds
- Ongoing support or retainer: 400 to 1,500 pounds a month once you've proven the first two
If you're wondering whether those numbers are too low or too high for your market, I've written a longer breakdown on how to price AI services for small business clients that covers what to charge once you're past the first few pilots. The short version: don't charge by the hour once you know what you're doing, charge by the outcome, but for your very first paid project, hourly or fixed-fee-low is fine because you're buying speed and proof, not margin.
Step 4: find clients where they already are, not where you wish they were
Forget building an audience from scratch. Go to:
- Local business networking groups (BNI chapters exist in most UK towns and meet weekly)
- Trade Facebook groups for the specific industry you picked, they're often more active than LinkedIn
- Your own past employer's supplier or client list, if there's no conflict of interest
- A direct email to twenty businesses in your chosen niche offering a free 20-minute audit call
Twenty cold emails, worded plainly, no jargon, gets you two or three calls. That's a realistic conversion rate, not a marketing failure. If you send twenty and get zero replies, the problem is almost always the subject line trying too hard to sound clever.
Step 5: run a proper first conversation, not a pitch
Your first call with any prospect should be mostly questions. What takes up your time that shouldn't? What have you already tried with AI that didn't work? Where does information get lost between you and your staff? You're listening for a specific, recurring, annoying task, not a grand strategic problem. Small businesses don't need a strategy. They need one thing fixed that's been bothering them for eight months.
If you want a structure for this that goes further than a chat, this is exactly what a paid audit looks like in practice, and I've laid out what should and shouldn't be included in one in this piece on what an AI audit for small business includes. Copy that structure for your own free discovery calls, it keeps you from rambling and keeps the client focused on decisions, not demos.
The uncomfortable part nobody puts in the "start your AI side hustle" articles
Here's the bit that gets left out. Most of your value in year one will not come from AI knowledge. It will come from you being organised, showing up on time, writing things down clearly, and following through on the boring stuff, the reminder email, the simple invoice, the one-page summary after the call. The AI tools themselves are commodities now. ChatGPT, Claude, Gemini, they're all good enough for 80% of small business use cases and the client can access any of them for twenty dollars a month without you.
What they're paying for is someone who will translate the tool into their specific, messy, human business. That's not a technical skill, it's a consulting skill, and it's the one most people trying to break into this space skip because they're too busy watching tutorials on the newest model release. I've seen people with far less AI knowledge than a self-taught hobbyist land 2,000-pound projects because they showed up prepared, asked good questions, and delivered a document that made sense. I've also seen skilled AI users lose deals because they couldn't explain a workflow in plain English to a 58-year-old shop owner. Don't skip the boring skills chasing the shiny ones.
Step 6: decide how much of a "business" this is
You don't need a company, a brand name, or a proper website to take your first three clients. You need an invoice template, a simple contract (a one-page scope of work protects you more than most people realise), and a business bank account once money starts moving, which in the UK you can open in an afternoon with most digital banks. Treat the first two or three months as paid learning, not a launch.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
If this is running alongside a full-time job, be honest with yourself about hours. Realistically you can manage one client's setup work and one discovery call a week without it eating your evenings entirely. Two clients a month at 500 pounds each is 12,000 pounds a year of side income for maybe six to eight hours a week, that's the honest range most people starting out should expect, not the "6-figure AI agency" numbers you see in ads.
Step 7: know when to specialise further, and when to stop calling it a "side" business
Around client five or six, patterns show up. You'll notice the same three requests keep coming from the same type of business. That's your signal to build a repeatable offer, a fixed package with a name, a fixed price, and a fixed set of deliverables, instead of custom-quoting every project from scratch. This is also the point where a lot of people realise they've quietly built something bigger than a side hustle, and it's worth reading through the wider list of side hustles you can start with AI to see how this compares to other options if you're weighing your time against other opportunities.
If you're running this entirely on your own, without a team, the constraints are different again, and I've written specifically about running a business of one with AI, which covers how solo consultants use AI tools to handle the admin side of their own business while they're out there fixing everyone else's.
What trips people up (from watching a lot of people try this)
Three patterns show up again and again. First, people spend money on a certification or course before they've spoken to a single potential client, when a free 20-minute call with a real business owner teaches you more than most paid courses. If you do want structured learning, I've broken down which training is worth paying for in this piece on AI training for business owners, most of it isn't, a handful of programmes are.
Second, people undersell the diagnosis and oversell the fix. They promise "AI transformation" in a first email, which sounds fake to a small business owner who's been burned by an overpriced website agency before. Specific beats impressive every time. "I can cut the time you spend writing follow-up emails by half" beats "I help businesses harness the power of AI" in every cold email test I've seen.
Third, and this is the one that stings, people quit after the first client says no. Two or three no's is normal, it's not a signal your idea is broken, it's a signal your pitch or your targeting needs a small adjustment, not a complete rethink.
When to bring in help, or become the help
If you're doing this well, at some point a client will ask whether they should just hire an agency instead of you, or whether one consultant like you is enough. That's a fair question and worth understanding from both sides, I've written about the difference between an AI consultant and an agency and which businesses need which, because knowing the answer for your clients also helps you position what you offer against the bigger firms circling the same small business market.
Starting this side business isn't complicated, but it is uncomfortable in the early stages because it involves cold outreach, undercharging on purpose, and doing unglamorous work for people who won't be impressed by jargon. That discomfort is the filter. Most people who try this quit in month one. The ones who stay past client three, the ones who keep the first case study honest and the first invoice sent on time, are the ones still doing this a year later with a client list built entirely from referrals.
Frequently asked questions
How much can you realistically earn from an AI consulting side business in the first year?
Most people starting part-time, alongside a full-time job, land somewhere between 8,000 and 15,000 pounds in year one from two to three regular clients paying 400 to 1,500 pounds a month plus occasional fixed-fee project work. Full-time consultants with a proven niche and referral network earn considerably more, but that's rarely the year one reality.
Do I need a certification to start an AI consulting business?
No. A certification helps with credibility on paper but does nothing for a small business owner who wants proof you can fix their specific problem. One solid case study from a real project will outperform a certificate on a website every time in the early stages.
What's the fastest way to get my first paying client?
Fix one small, specific problem for free or very cheaply for someone in your network, document the before and after with a real number, then send twenty direct emails to similar businesses offering a free 20-minute audit call built around that same problem.
Should I quit my job to do this full-time?
Not at first. Run it alongside your job until you have at least four or five paying clients and a repeatable offer. Most people who quit too early burn savings chasing a business model they haven't tested enough to know works for them.
Want this done for you? See how to build an AI workflow.
Related: looking for work with me on AI? Here is how I can help.
Related reading: AI Phone Receptionists for Small Business: What They Cost and Where They Quietly Lose You Money and AI Phone Agents for Small Business: What Happens When You Let AI Answer Your Calls.
Related reading: business owners hiring help often start with an AI consultancy in Israel.
