Do you need a website chatbot? What small businesses get wrong about AI chat
A competitor adds a chatbot to their website, so you add one too - a little bubble in the corner that says “Hi, how can I help?” Three months later, nothing about your enquiries has changed. You still do not know how many leads it has actually captured, and a few customers have mentioned it would not answer a simple question properly. This is the most common way small businesses get a website chatbot wrong: they buy it as the fix, when it is only ever a small part of one.
What a website chatbot actually does
At its best, a chatbot answers the questions you get asked twenty times a week - opening hours, what areas you service, roughly what a job costs - and captures a name and number from someone who would otherwise leave without filling in your contact form. That is a real, useful job. It is not a receptionist, it is not a salesperson, and it is not a system that understands your business. It is a script that recognises patterns in what people type and responds accordingly.
Where it goes wrong
The research on how people actually feel about AI chat is not encouraging. A 2024 Five9 survey of 4,000 consumers found 56% are often frustrated by AI customer service chatbots, and 48% do not trust the information the chatbots give them. Twilio’s 2025 research found 78% of people say being able to move from a bot to a real person matters to them, but only 15% say that handoff has ever felt smooth. And Omnisend’s 2025 survey found 39% of shoppers had abandoned a purchase outright because a chatbot interaction frustrated them.
None of that means a chatbot is a bad idea. It means most of the damage happens in a few predictable places:
- It tries to do too much. A bot set up to handle every question, including the ones that need judgement - a complaint, a tricky quote, a nervous first-time customer - frustrates people faster than no bot at all. It should hand off, not bluff.
- It has no way to hand anything off properly. A name and number typed into a chat window does not help you if it sits in a dashboard nobody checks, instead of landing with whoever needs to follow up.
- It is treated as the whole strategy. Owners install a chatbot hoping it will fix slow replies, missed calls, or a messy pipeline, when the chatbot is only ever handling the enquiries that arrive through one channel, in one format, well after the real problem started.
Do you need a website chatbot?
Usually, not first. If your actual problem is enquiries going cold because nobody replies fast enough, or jobs clashing because your booking details live in three places, a chatbot does not touch either of those - it just adds a fourth place to check. A chatbot is worth doing once you have a narrow, repetitive set of questions eating staff time, and somewhere reliable for whatever it captures to land. Installed before that, it is a cost that looks like progress without being any.
Why this fails without the data behind it
A chatbot that captures a lead perfectly still fails if that lead has nowhere good to go. If your customer records live in a spreadsheet and your job notes live in someone’s head, a chatbot just means the mess gets a name and number added to it faster. The fix is not a smarter bot - it is knowing where your enquiries actually need to land before you add anything that generates more of them. We go wider on this in Automating quoting, scheduling and after-hours enquiries for service businesses and Website enquiries going cold? Fixing slow lead response time before it costs you jobs.
Where to actually start
Before you add a chatbot, write down the five questions your customers actually ask most, and check where an enquiry currently goes when it lands - an inbox, a spreadsheet, a notebook by the phone. If you cannot answer both of those quickly, a chatbot is not your next move.
Our complimentary AI-Readiness Data Check takes about ten minutes and gives you a plain read on whether a chatbot would actually help your business, or whether something simpler would fix the real problem first.