Almost every business has been pitched an AI chatbot in the last two years. Some of them genuinely should have one. Many should not, and will waste money and irritate customers finding that out. This is an honest framework for telling which group you are in.
What an AI chatbot is now, versus what it used to be
The old generation ran on decision trees. You clicked “Billing”, then “Payment failed”, and got a canned paragraph. Anything unanticipated produced “Sorry, I didn’t understand that.” People hated them for good reason.
Modern assistants work differently. They read your actual website content, understand a question phrased in any way, and answer in natural language. Ask “do you work with businesses like mine” and a well-built assistant will handle it, where the old kind would not have had a button for it.
That distinction matters, because most scepticism about chatbots is based on the old experience. It is also why the failure modes have changed rather than disappeared.
When a chatbot genuinely earns its place
You answer the same questions constantly
If your inbox is full of the same handful of questions about pricing, availability, turnaround, coverage area or process, you are doing repetitive work a machine can do accurately. This is the clearest case, and the easiest to justify.
Your customers are active outside your working hours
An enquiry at 11pm typically waits until morning. Some of those people will have found a competitor by then. If your analytics show meaningful evening or weekend traffic, or you serve customers in other timezones, round-the-clock answers have real value.
You have a large catalogue or complex service range
When customers struggle to find the right product or service among many, a conversational interface often beats navigation. “I need something for X, budget around Y” is a faster route than filtering menus.
Your enquiries need qualifying before they are useful
If half of your enquiries are outside your budget range or service area, an assistant that gathers those details first will save you real hours. You get fewer but better conversations.
When a chatbot is the wrong answer
We would rather say this plainly than sell an integration that will not help.
Your enquiry volume is low
If you receive a handful of enquiries a week, a chatbot solves a problem you do not have. A clear FAQ page and a good contact form cost nothing and work fine. Automation is worth it when volume makes repetition expensive.
Your service is genuinely bespoke
If every project is scoped individually and pricing depends on a dozen variables, an assistant that guesses will create expectations you then have to walk back. Setting a wrong price expectation is worse than setting none.
Your website content is thin or out of date
An assistant answers from your content. If your service pages are vague, the assistant will be vague, and confidently so. Fix the content first. This is the most common reason chatbot projects disappoint, and it has nothing to do with the AI.
The real problem is that nobody answers the phone
If enquiries are being lost because responses are slow, a chatbot is a technological answer to an operational problem. It will paper over the gap without closing it.
The failure modes to design against
Confident wrong answers
The most serious risk. An assistant that invents a price, a delivery time or a policy creates a customer expectation you may be obliged to honour. Constrain it to your actual content, and instruct it to say it does not know rather than guess. Test this deliberately by asking questions your site does not answer.
No route to a human
Every conversation needs an obvious escape hatch. Customers who feel trapped in a loop leave angry, and they tell people. “Talk to a person” should always be one click away.
Aggressive pop-up behaviour
A widget that expands over content three seconds after arrival is a conversion problem, not a conversion aid. It is particularly damaging on mobile, where it can cover the entire screen. Let people open it when they want it.
Performance cost
Chat widgets are among the heavier third-party scripts. Loading one on every page affects responsiveness, which affects both experience and your Interaction to Next Paint. Load it after the page is usable, not before.
How to trial one without committing
- Collect your real questions first. Go through the last three months of enquiries and list what people actually asked. This tells you whether you have a repetition problem at all.
- Write the answers as content. Put them on your website as a genuine FAQ. This has value on its own and it becomes the assistant’s source material.
- Measure your baseline. Response time, enquiry volume, conversion rate. Without this you cannot tell whether the assistant helped.
- Start on one page. Usually the busiest service page. Not the whole site.
- Read the transcripts weekly. This is the step people skip and it is where the value is. You will find questions you never knew customers had.
- Judge it against the baseline after a month. If enquiry quality has not improved, remove it. That is a legitimate outcome.
The unexpected benefit
The transcripts are often worth more than the automation. They show you the words customers use, the objections they raise before contacting you, and the information they expected to find and could not. That is direct research into your own website’s gaps, and it usually improves your content more than the chatbot improves your response time.
What to take away
An AI assistant is worth it when you have genuine repetition, meaningful out-of-hours demand, a wide catalogue, or enquiries that need qualifying. It is not worth it when volume is low, when your service is highly bespoke, or when your website content is not solid enough to answer from.
Whichever camp you are in, the first step is the same: write down the questions you keep answering and put the answers on your website. If automation then makes sense, you have already done the hard part.
We build these where they help and say so when they will not, which is how we approach AI integration generally. If the honest answer is that your content needs work first, that is part of website development. Either way, tell us what you are trying to solve.
