"68% of businesses use AI"?
In practice: a chatbot

You have seen the percentages in the articles: “68% of small businesses already use AI”. If reading that made you feel left behind, relax. The gap between what people answer in questionnaires and what actually runs inside businesses is enormous. Knowing it helps you make better decisions than the people chasing the hype.

What the surveys actually measure

When a survey asks “do you use AI in your business?”, a “yes” covers everything: from the accountant who once asked ChatGPT about an email, to a system that enters orders on its own. Gene Marks, a consultant who has worked with thousands of small businesses for years, put it bluntly in Forbes: the surveys’ “using AI” usually means “monkeying around with a chatbot”. Real automation, where the system reconciles balances, places orders, answers customers and produces quotes, is “nowhere near happening” at small businesses yet.

In Greece the official numbers show the same picture. According to Eurostat, roughly 4 in 10 Greeks use generative AI tools in their personal lives, among the highest rates in the EU. At work? Fewer than 2 in 10. The same person who asks ChatGPT for Saturday’s recipe spends Monday entering invoices by hand.

Why you hesitate, and why you are partly right

The obstacles owners cite are always the same three: cost, reliability, trust. And they are not excuses. Large companies spend billions on AI and few see clear benefit in their results. Why should you take the risk, with far more limited resources, on something that has proven nothing to you?

I will not tell you to “be brave”. Your hesitation is a healthy defence against a market selling “AI transformation” with no deliverable. There are cases where you don’t even need a custom solution, and anyone who doesn’t tell you that is selling you something.

The difference that matters: a conversation tool or work getting done

A chatbot is a conversation tool. You ask, it answers, and the work remains yours. Useful, but it does not give you hours back.

Automation is a different thing: a system that executes, not one that chats. A customer’s voice message becomes an order in your system. The invoice arriving as a PDF gets transmitted to myDATA. The chat question gets answered from your own data, not from the internet. That is where the hours come back, and where the mistakes stop.

And since reliability is obstacle number two: proper automation keeps a human in control. Anything with risk goes through approval. The AI proposes, you decide. That is how trust, obstacle number three, gets built: not with promises, with small steps you watch working.

Where to start, without risk and without hype

1. Find the one task that eats the most hours of your month. Not ten, one. Usually it is data entry, copying between systems, or answers you type again and again.

2. Check whether it is worth automating: it repeats often, it has rules, it is stable. If not, leave it alone for now.

3. Start with a small first version you watch working before committing to anything bigger. If someone proposes automating “your whole business” in one go, keep your scepticism.

The 68% may be “using AI”. You don’t need to catch up with anyone. You need to solve one specific problem of your own, and to know honestly whether the tool is worth the money before you spend it.

If you want a second pair of eyes on where you lose time, write to me. I will also tell you if you need nothing at all, because that is an answer too.