The most common worry I hear from business owners isn't that AI won't work. It's that they won't be able to tell whether they're being sold something real, because they don't understand the technology well enough to argue.
That worry is misplaced, and I'd like to dismantle it. You don't need to understand how any of this works to evaluate a pitch. You need six questions, and the willingness to notice when one of them doesn't get answered.
1. "What will be different on a Tuesday?"
The most useful question there is. Ask them to describe an ordinary working day after this is installed, in terms of what specific people in your business will do differently.
A real answer sounds like: "Kelly stops re-entering job details into the accounting system, which is about four hours a week, and the entries stop having typos in them."
A pitch sounds like: "You'll unlock efficiencies across your operation and free your team for higher-value work."
The first one you can verify in a month. The second one can never be wrong, which is precisely the problem with it.
2. "What does it cost to run, not just to buy?"
Ask for the monthly figure after it's built: usage charges, hosting, and maintenance. Anyone who has built and kept one of these running will have a number, or at least a range, and will give it to you without flinching.
A vague answer here is the single most reliable warning sign I know. It usually means either they haven't operated one of these past launch, or the running cost is embarrassing.
3. "What happens when it gets something wrong?"
Not *if*. These systems get things wrong — they misread a figure, they answer a question confidently and incorrectly. That's a known property, not a defect, and it's manageable.
What you're listening for is whether they've thought about it. Good answers involve a person checking things above a certain value, or the system flagging when it isn't sure, or an easy way to see what it did and undo it.
If the answer is that it doesn't really get things wrong, they either haven't run one in anger or they're not telling you the truth. Either way you've learned what you needed to.
4. "Where does my information go, and who else can see it?"
You're entitled to a plain-English answer: which company processes your data, whether it's used to train anything, where it's stored, and what happens to it if you stop paying.
This is answerable in three sentences by anyone who has actually set one of these up. If it produces hedging or jargon, that's not a technology gap, it's a preparation gap.
5. "What would make you tell me not to do this?"
My favourite, because it's very hard to fake.
Anyone who has done this work a few times has a list — businesses too small for it to pay, processes nobody agrees on yet, records too scattered to build on. They'll answer readily, often with an example.
If the answer is that they can't think of anything, or that AI helps every business, you're talking to someone selling a category rather than solving a problem. There's no version of this technology that's right for everyone, and anyone who has genuinely deployed it knows that better than you do.
6. "Can I see one you've built that's still running?"
Not a demo. A demo is a controlled environment built to succeed. Something in daily use by a business that isn't them, ideally one you could ring.
The gap between building something impressive and keeping something useful running for two years is enormous, and it's where most of the actual skill lives.
The meta-signal: urgency
Running underneath all six is a simple test. If the pitch depends on you deciding quickly — a closing window, a competitor threat you hadn't noticed until they raised it, a price expiring Friday — the urgency is the product.
Genuinely useful tools don't need it, because the problem they fix was already irritating you before anyone rang. If you wait a fortnight and it still seems like a good idea when nobody is pressuring you, that's a much better basis for spending money.
Turning it around
You can point these at me too, and people should. What's different on a Tuesday, what it costs to run, what happens when it's wrong, where the data goes, when I'd tell you not to bother, and what's still running that I built. If I can't answer all six plainly about your specific situation, don't hire me for it.
That's the whole standard. It doesn't require you to understand a thing about how any of it works — only to notice whether the person selling it can explain what it will actually do for you, in words you'd use yourself.
