How Small Business Owners Can Start Using AI Without Getting Overwhelmed

You've seen the headlines. You've heard the promises. You’re probably dabbling with Chat GPT, Claude or Co-Pilot, but the gap between "AI could help" and knowing where to start feels enormous, and nobody seems to be explaining the bit in the middle.

Richard Geary and Scott Evans from AI Foundry joined us on What One Thing and instead of adding to the hype, they dismantled it. Richard has spent 25 years doing digital transformations in big global SaaS businesses, and the one thing he's learned applies to a company of 10 just as much as one of 10,000: always know your problem before you go looking for a solution.

If you're a small business owner trying to figure out where AI fits, without losing your mind in the process, here are some simple guidelines.

1. Start with the friction

The reason AI feels overwhelming is less about it feeling new and complicated and more about not having first defined what you need it to do.

Richard's seen the same pattern play out hundreds of times: a business buys a load of licences, hands them to the team, and waits for the productivity miracle. Of course, it doesn't come because people don't know what to use it for. The careful, conscientious members of your team won't touch it because they're worried it'll replace them or leak sensitive information. The rest are busy with their day-to-day work with no time to ‘play around’ with AI software. Six months later, you've spent a fortune on something nobody's using properly.

"We were a solution looking for a problem," Richard said. "And that for me was like a cardinal sin."

The fix is to flip it. Problem first. Tool second. Always.

Where to start: Write down the three things in your business that eat the most time or cost the most money every single week. Think repetitive, grinding, wish-someone-else-would-deal-with-it stuff rather than strategic deep work.

2. Your best people won't use it unless you give them permission.

This one was surprising. You'd think giving someone access to a tool is enough, but it isn't.

Richard and Scott have noticed that often, the people you most want using AI are often the ones who refuse to. And it’s not because they're resistant to change, but because they're conscientious. They're worried about training the model with company data. They're worried it'll make their role redundant. And they're worried they'll do something wrong.

"There's a permission thing, an approval thing that needs to happen from leadership to actually get AI tools embedded into the way that people work," Richard said.

Without that, your keenest self-starters will play with it in isolation and everyone else will pretend it doesn't exist. You get pockets of improvement instead of a big shift.

Where to start: Have a conversation with your team before you hand out a single login. Tell them why you're exploring it, what it's for, and what it isn't for. Make it clear that experimenting is encouraged and that nobody's job is on the line because of a chatbot.

3. AI is the first technology where experimentation is the whole point.

Every other technology rollout in the history of business has followed a script: here's the tool, here's the process, here's the expected result. AI doesn't work like that.

Richard put it simply: "AI is the first technology where we provide access and then we encourage experimentation. And that's very different to any other technology you've ever rolled out."

Two people with access to the same tool will use it in completely different ways, and both can be right. A fleet of van drivers needs route optimisation and smarter loading. A team of plumbers needs faster quoting. A team of accountants needs help keeping up with legislation changes. Same technology, completely different applications, and none of them start with "buy Copilot and see what happens."

Where to start: Pick one person in your team, the curious one, the one who's already poking around with ChatGPT at home. Give them a paid licence and a specific problem from your list. Let them experiment for two weeks and report back on what they found. That's your proof of concept.

4. A cup of coffee and £20 a month can replace a £10,000 consultant.

Phil put a scenario on the table that’s highly relatable for small business owners. Bob runs a company with 20 staff. He knows AI could help but doesn't want to drop £15,000 on a consultant, and even if he did, he wouldn't know what to brief them on. So what does Bob do on Monday morning?

Richard's answer was the most practical thing in the whole episode: spend £20 a month on an AI tool, any of them, and don't start by asking it questions or telling it to ask you questions.

Tell it: you are an expert business consultant, your job is to help me figure out where AI can impact my business. Then talk to it about your team, your customers, your frustrations, your bottlenecks. Let it interview you. Richard reckons two days of doing this will give you more clarity than a five-figure consultant engagement.

"The most important thing in AI is context," he said. "If you sit in isolation and try and think of all the things your business does, you'll miss 40% of what you do."

Phil took it a step further: get your staff to do it too. Have each person sit down and let the AI ask them about their role, their processes, what takes too long, what breaks. By the end, the tool has a more complete picture of your business than you've probably ever given anyone.

Where to start: Block out two hours this week. Open ChatGPT, Claude, Gemini, whichever. Type: "You are an expert business consultant. You are going to help me figure out where AI can impact my business. Interview me." Then just answer honestly.

5. When everything feels urgent, filter by time and money.

Hayley asked the question a lot of business owners are pondering: when you've got a hundred things that could be better, how do you pick which one to start with? Because if you try to fix everything, AI just becomes another item on the to-do list that makes you feel behind.

Richard's filter was disarmingly simple: time and money. What's costing you the most of either? What would free up time to go and do something that generates more money? Start there. Doesn’t have to be complicated or sound good on LinkedIn. It should just be the one that's bleeding you dry, week after week.

And here's the part that surprised Hayley: you can use the AI tool itself to help you prioritise. After you've spent a few hours giving it context about your business, it can rank your problems against whatever measure matters most to you. Staff retention. Revenue. Operational cost. You choose the lens; it does the sorting.

"Most people work in their business, not on their business," Richard said. Spending even eight hours over a month talking to an AI tool about what your business really does is eight hours of working on your business. For most small business owners, that's more strategic thinking time than they get in a quarter.

Where to start: After your first interview session, ask the tool: "Based on everything I've told you, what are the three biggest problems in my business ranked by the impact on time and money?" See what comes back. You might be surprised at what it spots that you've been too close to see.

6. Don't trust it. Challenge it like you would a stranger.

The trust question always comes up, and Scott gave the most honest answer possible: yes and no. You can trust it if you've given it good context, asked specific questions, and been precise about what you need. You can't trust it if you've been vague and accepted the first answer without pushing back.

Richard went further: treat it like a consultant you've never met. You wouldn't let a stranger walk in and restructure your business on day one. So don't let an AI tool do it either.

His approach: demand multiple sources, challenge every recommendation, and never accept a single-source answer. Push back until you're satisfied the output is grounded in your reality and evidence. "You can be really rude to AI," he said. "It doesn’t care. It's not going to take it personally."

He also explained AI hallucination: when AI doesn't have enough information, it fills in the gaps. Confidently. That's where the danger is, not in the tool being wrong, but in it being wrong with conviction. The fix is simple: keep giving it context, keep questioning the output, and if something feels off, make it show its working.

Where to start: Every time the AI gives you a recommendation, ask "why?" and "what are you basing that on?" If it can only point to one source, or if the reasoning feels thin, push it harder. Treat the first answer as a draft, not a decision.

The bigger picture

Five years ago, getting this kind of strategic clarity about your business meant a big project: expensive consultants, months of scoping, and a solid chance you'd end up trying to fix 50 things instead of five.

Now it's £20 a month, a coffee, and the willingness to be honest about what's not working. Richard runs a one-person business that operates like a team of four or five, spending around £220 a month on AI tools. AI isn’t a magical cure-all, but the barrier to getting started has dropped so low there's no good reason to keep putting it off.

The tools don't do the thinking for you, but they'll think with you, if you give them something real to work with.

Richard Geary and Scott Evans run AI Foundry. You can find them at wearefoundry.ai or connect with them on LinkedIn. Their free community is open to anyone who wants to learn about AI without feeling stupid for asking questions.

Brought to you by Affirm IT Services Ltd and Corbar Accounting.