I grew up in a circus, so I feel qualified to talk about the circus that is AI right now. An AI agent hacked a government website last month. Every second LinkedIn post is about it. Aussies are fully obsessed.
If you're not using AI, I respect that. If you are, I still respect you, but you have to think about how. Because your customers are about to start asking, and from what I’ve seen,the sharper ones already are.
Here are the four questions I think your small business will hear in the next twelve months, and how to have a simple answer ready for each.
1. What are your rules for using AI?
Customers want to know whether you use AI, how you use it, and whether they can still rely on you for quality. Some businesses use these tools well. Others let the tool do the job and forget to check that the thing the customer is paying for actually got done.
I run an ISO certification body, and we had to work this out for ourselves. How would our clients feel if we used AI during an audit? Or to help write a report? The surprise was that the hardest conversation wasn't with customers. It was with our own team, agreeing on what's fine and what isn't.
So write a one-page rule book. Which tools are you okay to use? What can go into them? Which outputs get a mandatory human check before they reach a customer? And what do you do when a tool does something you didn't expect?
Decide now what never goes in: client personal information, passwords, financials, the fact that the boss likes pineapple on pizza, contracts, you get the drift. Remember, this applies unless you have a proper contract with the tool, run a model yourself, or have some other way of managing that risk.
Then get your team to agree to it. A rule book nobody signed up to is just a poster.
2. How do we know your AI tools are safe?
This one is about trust. How did you test the tool? How do you know it can do the job? And how do we know you actually did those tests?
You need a way to show this, not just say it. For some businesses that's a formal certification like ISO 42001. For 95% of organisations, it's simpler: a short example of how you tested a tool before you trusted it (run it on a job you've already done by hand and compare the two), or the step where you strip out identifying details before anything goes in. Keep it plain. Why you use AI, how you use it, and what you do to use it safely.
Here's a test I use. If you wouldn't feel comfortable telling a customer how you're using AI, that's your
answer. Fix the use, or stop. A customer who finds out later and doesn't like it won't be a customer for long.
3. What do you do if something goes wrong?
A quick aside first. I know some of you are about to paste this article into ChatGPT and ask it to write your answers to all four questions so you've got something to show customers. Please don't. Really think about why you're answering these questions in the first place.
Now, back to question 3:the plan for when things go wrong. Write down who gets called, how you stop the tool, and what you say to the customer. Yes, prepare the apology/problem solving before you need it. Keep the whole thing to one page. Nobody reads a twenty-page incident plan.
Find out whether your business has a legal duty to report certain data breaches. Many small businesses don't. If yours does, write the steps down.
Then do a practice run with your team, and make it safe to put your hand up. The people trying to be innovative are the ones most likely to stuff something up. You want them telling you, not hiding it.
4. Can your team still do their jobs without AI?
This is the most under-discussed question in AI right now. Everyone is focused on efficiency, headcount and profit. Almost nobody is asking what happens when the team has leant on the tool so long they've forgotten how to do the thing themselves.
Your customers are choosing a team, not a subscription. They want to know your people are still being trained, still getting better, still the ones they'd pick. So decide what your team must be able to do without AI, and keep them doing it. Pick one task a month to do the old way, and treat it like training, because it is.
Where to from here
There are good frameworks out there to guide you on all of this. But the real point is simpler. Your customers are using AI. You're using AI. The businesses that think about how they use it, and talk openly about it, will be trusted more than the ones that use it and pretend they don't.
Have I said AI too much?