What are chatbots, and when it makes sense to use one for customer service

Anyone searching "what are chatbots" has almost always already read a definition somewhere, and doesn't quite remember it because it sounds like a technical term used to sell something. The useful question, if you run a business, isn't what a chatbot is in the abstract: it's what would actually change if you put one in front of your customers, and what to watch out for before you do.

What it actually does

A chatbot is a program that answers messages in your place, on your website, on WhatsApp or on social media, without a person online at that moment. There are two kinds, and it's worth knowing the difference before choosing one.

The first kind follows rules written in advance: for a frequent question ("what are your hours", "how do I book") it replies with a text that's already prepared, and outside those questions it has nothing to say. The second kind, built on generative AI, writes the answer on the spot by reading the message: it covers more cases, but it can also produce a wrong answer with the same confidence as a correct one, because it doesn't "know" it's wrong, it generates plausible text. That's not a technical detail: it decides how much you can trust it to answer unsupervised.

How many businesses are already using it

A useful reference point before deciding: according to ISTAT's Imprese e ICT, anno 2025 report, published on 15 December 2025 (data checked on 2 August 2026), 16.4% of Italian firms with at least 10 employees use at least one AI technology, up from 8.2% in 2024. The jump is sharper among large companies, from 32.5% to 53.1%, while among small and medium ones the figure doubled, from 7.7% to 15.7%.

It's just as important to say what these numbers don't show: ISTAT doesn't isolate a specific figure for "chatbots", it counts them inside broader categories such as text generation or written-language processing. Anyone claiming that a certain share of businesses "already uses a chatbot" is rounding off a number that, broken down this way, doesn't currently exist in an official source. What the data does say with certainty is the direction: AI adoption among Italian businesses, of any kind, is growing fast, and the gap between small and large companies remains wide but is narrowing.

The obligation that starts today, not years from now

There's a reason to talk about this right now, and it isn't an editorial trick: as of 2 August 2026, that is, today, Article 50 of the EU's AI Act is fully applicable, the one covering transparency toward anyone interacting with an AI system.

The substance, without the technicalities: anyone who puts a customer-service chatbot online must make sure whoever is typing knows they're talking to an AI, unless that's already obvious from the context. The duty to design that notice formally falls on whoever builds the system, but if you have it custom-built or put it under your own brand, you become the one responsible for it, and either way it's on you, in practice, to check that the notice is actually visible on your site, not hidden in a setting nobody ever opens. There's no exemption for small businesses from the obligation itself: only the way penalties are calculated changes in case of a breach, not the requirement to comply.

When it makes sense, and when it doesn't

A chatbot pays off when the questions a business gets repeat, are few, and are always the same: hours, address, availability, how to book. In that case a system that answers immediately, even at night or on a Sunday, takes work off your plate without taking away quality: the answer was already the same one, you were just the one giving it, with more effort.

It pays off less, or risks making things worse, when customers' questions genuinely differ from one another, or when a wrong answer is costly: a customer deciding whether to book for a week, or whether to trust you with an important job, won't forgive a chatbot that gives them inaccurate information in a confident tone. In those cases the honest choice isn't to avoid AI altogether, but to keep it on the simple questions and route everything else quickly to an actual person: the risk isn't using a chatbot, it's leaving it in charge of cases it shouldn't handle alone.

How to decide, in practice

The question that matters isn't "do I need a chatbot", it's more concrete: how many times a week do you, or someone on your team, answer exactly the same three or four questions? If the answer is "often", automating that part frees up real time, not just time on paper. If the answer is "rarely, every customer asks something different", a chatbot doesn't solve a problem you don't have: it adds one, keeping it up to date.

I build custom automation and chatbots for small businesses starting from exactly this question, not from the technology: we map out what genuinely repeats in your business, and automate that, leaving people to do what should stay theirs. The cost is worked out case by case, because it depends on how many questions it needs to cover and where it needs to answer, not on a one-size-fits-all price list.