What AI can and cannot do in a business

27 September 2026·Auttomix Team

What it takes on today, and where people are still needed

According to ELSTAT, the Greek statistics authority, one in two Greeks aged 16 to 74 were already using AI tools in early 2025. Most of them, though, used them for personal reasons and not for their work.

And as use grows, so does worry. In an international Pew Research Center survey, Greece is among the countries most worried about losing jobs to AI.

So for you as a business owner, the question is not whether AI is good or bad. It is more practical: which jobs in your business can it take on, and which should stay with people?

What can AI do in a business?

Here things are clear. AI does best with work that repeats, follows clear rules and comes in large volumes: answering the same questions, sorting requests, taking notes, drafting texts, entering data.

Three real examples, from three very different sectors:

In a hospital. In the emergency department of St George's Hospital in London, an AI tool takes notes while the doctor talks to the patient. In the pilot, medical and nursing staff saved an average of 47 minutes per shift, enough for each of them to see one more patient. A family doctor who tried a different AI tool from the UK health service put it simply: "It hasn't replaced our judgement – it's given us back the time to use it."

In HR. At IBM, employees' everyday requests to human resources go first to AskHR, an internal AI agent. According to the company, it answers 94% of common questions on its own, such as leave or certificates. People in the department take the more complex cases.

In the factory. Ford uses AI in quality control on its production lines. Here, though, the story has another side, which we will come to below.

And in Greece? Unfortunately, examples are few to none. We searched for this article and did not find a single Greek small or medium business describing publicly, in its own words, what work it has handed to AI. That fits what almost every survey shows: Greek businesses are adopting AI very slowly.

If you want to see how this works in practice on a smaller scale, in hotels, clinics and offices, our first article on AI in business has everyday examples. And if you want a process like this set up for your own business, from start to finish, that is what Custom AI Agentic Workflows are for.

What can't AI do?

This is where honesty is needed, because this part is missing from most presentations.

It doesn't know what nobody told it. Ford believed that giving the AI its design specifications would be enough to produce a quality product. It wasn't. Many of its most experienced engineers had already left, and their knowledge had never made it into the system. So Ford hired about 300 engineers with years of experience. Charles Poon, the company's vice president of vehicle hardware engineering, explained: "Artificial intelligence is a fantastic tool, but it's only as good as the information you use to train it." Today these engineers review designs every week, before they reach production, and help improve the data the AI is trained on.

It can say something wrong with complete confidence. OpenAI itself, the company behind ChatGPT, explains that language models "hallucinate" answers because the way they are trained rewards guessing more than saying "I don't know". That is why anything that reaches a customer needs limits and checks.

It doesn't take the blame. When an airline's chatbot gave a customer wrong information about a refund, a Canadian tribunal ruled in 2024 that the company was responsible, not the chatbot: "It makes no difference whether the information comes from a static page or a chatbot." Whatever AI tells your customers, you are the one saying it.

It can't judge the exceptions. The customer asking for something outside the rules, the complaint that needs a careful touch, the decision that costs money if it goes wrong. That is where a well-built AI assistant stops and passes the case to someone on your team, with all the details.

Which jobs should you hand to AI and which should you keep?

A simple way to sort it out: take one task in your business and see which of the three categories it belongs to.

Most tasks in a small business do not sit entirely in one category. Usually part of them moves into the middle one, and that is where the biggest benefit is.

Will AI take our jobs?

The worry makes sense, but the evidence shows something more complicated than the headlines.

In last year's global McKinsey survey, many businesses using AI expected to cut staff. In this year's, far fewer said their staff had actually shrunk. Most saw almost no change in the number of people they employ.

In fact, some of the companies that use AI most are hiring more young people. IBM, the same company that lets AskHR answer common questions, announced that in 2026 it will triple entry-level hiring in the US. Its chief HR officer explained: "If we don't continue to invest in entry-level hires, what happens in 3–5 years?"

So what changes is not so much the number of people as what they do. AI takes on the repetitive work and frees your team for what really matters: the customer, the decisions, the relationships.

AI in Greece: how far behind are we?

Quite far. According to Microsoft, in June 2026 the share of people using AI tools in the United Arab Emirates was more than three times that in Greece. Singapore, South Korea, the US and most of Europe, Italy and Portugal among them, were also ahead of us.

For a business, that has a cost. A competitor that has already handed its repetitive work to AI gains time for its customers, with the same team. The longer you wait, the more ground you will have to make up.

What to know before you start

One obligation few people know about, from the EU AI Act: since 2 February 2025, every business that uses AI systems must take steps so that its staff learn how they work. The regulation does not require exams or a specific level of knowledge. In practice, your team needs to know what the tool does and where it can go wrong.

And a practical tip: start with one task, not the whole business. Pick the one that eats most of your time each week, as long as it belongs to the first or second category in the table. If it works there, you will know what the next step is.

---

If you would like to look together at which tasks in your business AI can take on and which should stay with people, let's talk.

Auttomix Team

← All articles