An AI Chatbot Is the Last Thing Most Business Websites Need

An AI chatbot is the last thing most business websites need, and it is the first thing a good chunk of the industry is now trying to sell them. The pitch is everywhere, and it is about to get louder. So I want to make the unfashionable case plainly, from the position of someone who runs the infrastructure under a lot of small business websites and watches what actually happens on them.

Let me head off the obvious objection first. I am not anti-AI. I built a platform whose whole premise is that AI has finally made it sensible for an owner to build their own professional site. That is real, and I will come back to it. But there is a world of difference between AI that builds the thing and AI you bolt onto the front of it to talk to your customers. One has earned its place. The other, for most businesses reading this, has not. Not yet, and possibly not ever.

Where the pressure is coming from

The pressure is not really about chatbots. It is about fear, and the fear is well documented.

A Google survey of four hundred Irish SMEs, carried out with Amárach Research and reported by RTÉ in March, found that the single biggest barrier to adopting AI was not cost and not skills. It was fear of making mistakes, cited by around three in ten owners, ahead of a lack of skills at roughly twenty-seven per cent and cost at about twenty-four. More than half of those owners, fifty-seven per cent, said they felt behind their competitors.

A separate report from Viatel and Amárach last September, the AI Horizons study, put security worries near thirty-eight per cent and a lack of technical expertise around thirty-five at the top of its list, with close to a third of firms not using AI at all. Small samples, both of them, a few hundred owners between them, so take the precise decimals with the usual pinch of salt. The direction of travel is not in doubt though.

Tellingly, in that same Viatel study almost every owner already using AI found it genuinely useful, while the large majority had no plan or policy guiding where they used it. Useful, but unguided. That gap is exactly the one a vendor fills for you, in their favour rather than yours.

Read those numbers as a salesperson and you see an opportunity. Here is a market of anxious owners who believe they are falling behind and are not sure where to start. The easiest thing in the world is to sell that person a single, visible, bolt-on product that makes them feel caught up. A chatbot fits the brief perfectly. It sits in the corner of the screen, it looks modern, and it lets an owner tell themselves they have done AI. Whether it does anything useful is a separate question, and not one the pitch tends to dwell on.

What a chatbot actually is, and what it is not

Strip the branding away and a website chatbot is a language model wired to some text about your business, and sometimes not even that. So what does that mean in practice? It means the thing answers confidently whether or not the answer is correct. A person who does not know hesitates, or says they will check. The model does not. It produces a fluent, plausible reply, and a wrong one looks exactly like a right one. That reply then reaches your customer, in your name, on your site, with your logo sitting above it.

I will be honest about my own misjudgement here. When support bots first appeared, I assumed they would take work off our team. I had it backwards. The good ones need watching, and the bad ones generate more work than they remove, because every confident mistake becomes a complaint you then have to clean up.

Air Canada found this out in a tribunal, and the story is worth knowing before you deploy anything similar. Its website chatbot told a grieving customer he could book a full-fare flight and claim a bereavement discount afterwards. He could not; that was never the airline's policy. When he took it to the British Columbia Civil Resolution Tribunal, Air Canada argued, remarkably, that the chatbot was a separate entity responsible for its own words. The tribunal did not entertain it and held the airline liable, because a business is responsible for the information on its own website, whoever, or whatever, put it there. As the law firm Pinsent Masons noted at the time, that principle is not unique to Canada.

Picture the smaller, quieter version of that on your own site. You set the bot up in a busy week, train it on last year's information, and move on, because you have a business to run. Three weeks later it is cheerfully telling a customer you open on the bank holiday Monday when you are closed, or quoting a price you changed in January. You have no idea it is happening, because the whole appeal of the thing was that you would not have to think about it. By the time anyone notices, the lead is gone and the review is already written.

What your website actually needs first

Here is the part that gets lost. Most of the problems an owner hopes a chatbot will solve are not chatbot problems at all. They are website problems wearing a disguise.

When a visitor cannot find your opening hours, the answer is to put your opening hours on the page, in plain sight, not to make them open a chat window and interrogate a robot for something you could have shown them outright. If people keep ringing to ask whether you do a particular service, the fix is a clear services section and a visible phone number, not an assistant that paraphrases the same answer less reliably.

A fast site, a contact number above the fold, the three questions everyone actually asks answered directly on the page, a form that works on a phone. Get those right and you remove most of the demand the chatbot was supposed to absorb in the first place.

This is where AI genuinely does pull its weight, and it is the part worth paying attention to. The hard bit of a website was never the chat widget. It was building a fast, well-structured site at all, the thing that for twenty years needed an agency and a few thousand euro. That barrier is the one that has actually fallen. You can now describe your business in plain English and have a professional WordPress site assembled in about a minute, properly laid out, quick on mobile, with the contact details and the key answers where customers can see them.

WordPress runs close to forty-three per cent of the web, according to W3Techs, precisely because it does that unglamorous structural work well. That is AI earning its keep: building the asset, not chattering on top of it. For the wider picture of how that shift is reshaping things, I have written about what AI-powered WordPress actually means for a business owner.

When a chatbot does earn its place

Now the honest concession, because there is a real case for these tools and pretending otherwise would be the kind of one-sided pitch I am complaining about.

If your business genuinely fields the same narrow set of questions hundreds of times a week, a well-built assistant can be a real help. Think of a Limerick accountancy firm in the weeks before the tax deadline, answering the same three questions about filing dates and what documents to send, over and over, while the actual work piles up. In a case like that, an assistant trained tightly on your own verified information, with a clear path to a human the moment it is out of its depth, can take a real load off the phones.

The deciding factors are volume and repetition. A high volume of nearly identical questions, paired with a willingness to supervise the thing, is what separates a chatbot that pays for itself from one that just sits there looking modern.

And supervise it you must, which is the limit nobody selling these likes to mention. Even a well-trained assistant will get something wrong eventually, and the moment it does, that answer is yours. So a bot is not something you deploy and forget. It is something you set up, then verify regularly, correct when it drifts, and retrain whenever your prices or services change. That is ongoing work. If you do not have the appetite for it, you do not have the appetite for a chatbot, and that is a perfectly sensible place to land.

The AI worth your attention

Step back and the priority order is obvious. The AI that changes a small business is the one that gets a real, fast, well-built website live in the first place and keeps it cheap to run. A chatbot is a niche tool for a specific volume problem, not a rite of passage everyone has to perform to prove they are keeping up.

There is, by the way, another flavour of AI worth being just as careful about: the kind that writes the words on your pages rather than answering questions in a window. I have argued elsewhere that one kind of AI on your site pays for itself while another quietly costs you customers, and the same rule of thumb covers both. Let AI do the structural work no customer ever sees, and keep a human hand on anything a customer actually reads or is told.

Get the order right

So before you add an assistant to your site, get the order of operations right. Make the website itself do its basic jobs brilliantly: load fast, show your number, answer the obvious questions, work on a phone. Most owners who do that find the urge to bolt on a chatbot quietly fades, because the demand it was meant to soak up was really a symptom of a site that was not answering for itself.

If you then find you genuinely have a high volume of repetitive questions and the appetite to keep an assistant honest, trial one by all means, eyes open, supervised, and trained on what you know to be true. That is a real decision with a real upside for a small number of businesses. For everyone else, the modern, useful, money-saving form of AI is the one that built the site in the first place. Get that part right, and the question of whether you need a chatbot mostly answers itself.

Sources

• RTÉ News, One-fifth of Irish SMEs not using AI as yet (March 2026), the Google and Amárach Research survey of 400 Irish SMEs used for the barriers and competitor-anxiety figures
• ThinkBusiness, Are SMEs embracing AI correctly?, coverage of the Viatel and Amárach AI Horizons study used for the security, expertise and adoption figures
• Pinsent Masons (Out-Law), Air Canada chatbot case highlights AI liability risks, on the tribunal holding a business liable for its website chatbot's answers
• W3Techs, Usage statistics of WordPress, the source for WordPress's share of the web