Skip to contact
Seamwork AI
Book a free call
← Blog

Is a chatbot a good first AI project?

Usually not. For most small businesses, a chatbot on the website is one of the riskier places to start with AI, and it rarely saves the time you're actually losing.

A chatbot is a small chat window on your website that answers visitors' questions automatically, like a receptionist who never goes home but only knows what's been written down for them. There are good reasons for having one, and we'll cover those below. But the evidence suggests it's usually better as a second or third project than a first.

How do customers feel about AI customer service?

Of everything people use AI for, customer service is where they're most often disappointed. Qualtrics XM Institute surveyed more than 20,000 consumers across 14 countries, including the UK, in the third quarter of 2025. Nearly one in five of those who had used AI for customer service got no benefit from it at all, a failure rate almost four times higher than for AI use in general. Asked to rate AI uses on convenience, time saved and usefulness, consumers put customer service near the bottom.

If a chatbot is top of your AI list, that's worth bearing in mind. It's also why we'd suggest starting somewhere else.

Why a chatbot is a risky first AI project

A chatbot talks to your customers, so any mistake it makes happens in public. That makes it a risky place to try a new technology for the first time. Errors happen in front of a potential customer, at the moment they're deciding whether you seem reliable, and most of the time you'll never know.

Which? tested the online chat of seven large retailers (Amazon, AO, Apple, Argos, Currys, John Lewis and Very) over March and April 2023, asking for help with faulty electrical goods. All seven gave inconsistent or incorrect advice on consumers' legal rights. AO and Very offered only a chatbot. Apple, as both retailer and manufacturer, wrongly stated a 14-day return window on an iPhone.

That was 2023, and the tools have improved a great deal since. The underlying risk is the same, though. Those are large companies with teams dedicated to getting this right. If their online chat gave customers wrong information about their rights, it's reasonable to expect a small firm to find it hard to get right first time.

A chatbot only knows what you've written down

A chatbot can only answer from what it can read, and this is the part that's easy to overlook. Most small businesses don't have their answers written down. There's a website written a couple of years ago, a price list that's slightly out of date, terms and conditions, and everything else lives in the owner's head, in emails, or in a WhatsApp group. A chatbot pointed at that will answer confidently from the thin, out-of-date material it can see, and make up the rest.

So most of the real work in a chatbot project isn't the chatbot. It's writing down, perhaps for the first time, your actual lead times, what you do and don't cover, what happens when something goes wrong, and what you charge. That's useful work in its own right. It's also often the point where a clear page of answers on your website turns out to solve most of the problem by itself, at no cost.

Will a chatbot save you much time?

Probably less than you'd hope, at small volumes. Automation pays off when there's a lot of the same work, so it's worth running your own numbers first. If your website brings in five enquiries a day and each takes four minutes, that's twenty minutes a day. A chatbot that handled every one perfectly would save under two hours a week. In practice it won't handle all of them, because the awkward enquiries are exactly the ones it struggles with. Saving the easy half of twenty minutes a day is hard to justify as a business case.

Do customers want to talk to a person instead?

They say they do, though what they actually do is more mixed.

YouGov surveyed 2,046 UK adults between 5 and 8 June 2026, on a nationally representative sample. Asked whether they'd rather deal with a real person or an AI system when contacting a business, 90% preferred a real person, 74% strongly, against 2% who preferred AI. Two caveats apply. The research was commissioned by Answer4u, a telephone answering company with an obvious interest in the result, and the question was about contacting a business in general, not website chat in particular.

The broader concern is better supported. In the Qualtrics research, half of consumers worried that a company's use of AI would stop them reaching a human, and 53% named misuse of their personal data as their top concern about automated interactions, up eight points in a year.

On the other hand, people clearly do use chatbots. The ONS Opinions and Lifestyle Survey found in May 2023 that 34% of adults in Great Britain had used an AI chatbot in the previous month, half of them for customer service, rising to 63% among chatbot users aged 70 and over. What people say they prefer and what they actually do aren't the same, so it would be overstating things to say customers simply dislike chatbots.

The clearest problem is for people with disabilities. Which? surveyed 732 disabled consumers through the Research Institute for Disabled Consumers in July and August 2023: 65% of those who used a chatbot didn't find it an accessible experience.

Where is the time really going?

For most owners, this is the deciding point. Evenings don't usually disappear answering the same website question. They go on quoting, invoicing, chasing payments, scheduling, writing up jobs, and typing the same information into three systems that don't talk to each other. A chatbot doesn't help with any of that. It sits on the one part of the business that was usually already working reasonably well.

The Institute of Customer Service found in March 2026 that only 24% of employees believe their organisation has seen a return on its AI investment so far. That isn't an argument against AI, which can be genuinely useful. It's a reminder that where you point it matters.

If the admin is what's eating your evenings, we're happy to look at automating the quoting, invoicing or chasing that takes up the most time.

What's a better first AI project?

Choose work where a mistake gets caught before it reaches a customer. If an internal tool gets something wrong, someone says "that's not right" and fixes it. If a customer-facing one gets it wrong, you may lose an enquiry and never hear about it. It's the same AI and the same error rate, with very different consequences. Until you've seen how AI handles your own material on real work for a few weeks, it's sensible to start where mistakes are cheap.

Four questions help pick the first job:

  • Is it repetitive, with the same steps every time?
  • Is there enough of it that saving ten minutes matters?
  • Is it clearly defined, with a right answer someone could check?
  • Is it currently being done in the evening?

Anything that gets a yes to all four is a better first project than a chatbot. That often means drafting quotes from a standard specification, turning rough site notes into a written report, pulling figures off supplier invoices, first drafts of repetitive tender answers, or collecting the same three documents from every new client without writing the email yourself each time.

It's also what businesses already using AI mostly do. The ONS reports that improving business operations is the most common use of AI in every business size band, at close to 60% of firms using AI, well ahead of anything customer-facing. For more on that, see our article on how many UK businesses actually use AI.

When is a chatbot the right choice?

There are real cases where a chatbot makes good sense.

You get a lot of the same enquiries. If you're answering dozens of the same questions a day, the numbers work the other way.

Your answers are already written down. If you have an up-to-date help page, published policies or a proper FAQ, most of the hard work is done, and the risk drops sharply because the chatbot has something reliable to answer from.

Replying out of hours wins the job. Emergency trades, hospitality bookings, or anything where the customer works down a list until someone replies. Losing a job at 9pm because nobody answered is worse than giving a good-enough answer at 9pm.

Even then, it should hand over to a person quickly, say clearly that it isn't a person, and never guess about anything involving money or someone's legal rights. If your business fits one of these cases, we're happy to build a chatbot that works that way.

What should you do next?

If you have a chatbot in mind, start with the writing. Draft the twenty answers it would need. That's an afternoon's work and costs nothing. Then ask whether putting those answers on your website as a clear FAQ, with your phone number at the top, solves the problem. If it does, that's a good result, not a failed project. If it doesn't, you now have exactly the material a chatbot would need, and you can decide based on evidence.

If your enquiries are manageable and it's the admin that's taking your evenings, start there instead. It's less visible, but it's where the hours are. You can read more about what we do if you'd like a hand with that.

Sources

Checked on 16 October 2026, including a search for newer waves of the Qualtrics and Institute of Customer Service research (none found). Consumer attitudes to AI are moving quickly and several of these surveys are annual, so check for a newer wave before quoting a figure back at anyone.

The next step

The cost of waiting is not zero.

Somewhere in your market, a business the same size as yours has stopped writing follow-ups by hand. They are quoting faster than you, chasing money you are still chasing manually, and starting Monday already knowing where they stand. That gap does not stay the same size.

Half an hour on the phone. You describe your week, we tell you honestly which parts of it are worth automating and which are cheaper left alone. No pitch, and nothing to sign - you decide whether you want to take it further.

Book a free call

Or just ring us on 07816 970 768, or email hello@seamwork.co.uk.

We reply within a working day to have a chat about what we can provide and if we cannot help you we will say so, no pressure, no commitments, we just love showing off cool tech :)

We explain what we do with this in our privacy notice.