When a chatbot genuinely helps
Three conditions, and it needs all three to be worth building.
- Visitors repeatedly ask a small set of questions. If your inbox shows the same eight questions, that is a chatbot. If every enquiry is different, it is not.
- The answers exist in writing and are stable. Delivery, sizing, compatibility, opening hours, order status. Something the model can be grounded in rather than guessing at.
- Answering them is currently costing somebody real time. If it is ten minutes a week, a better FAQ page is the answer and it costs nothing.
The strongest case is e-commerce with a large catalogue, where "does this fit", "when will it arrive" and "can I return it" account for most contacts and the answers are all in the product data already. The weakest case is a five-page brochure site, where a chatbot is a widget that covers the phone number.
The specific ways chatbots annoy people
- Opening themselves. An unprompted panel over the content somebody came to read is an interruption, and it fires on the visitor who was already about to buy.
- Hiding the human. If the route to a person is buried behind three attempts, the chatbot has become a cost-cutting wall and customers describe it that way.
- Answering confidently and wrongly. An ungrounded model will invent a delivery time or a returns window, and you are accountable for what it said.
- Forgetting. Asking for an order number that was given two messages ago is the fastest way to lose somebody.
- Pretending to be human. It always comes out, and it costs more goodwill than the honesty would have.
What it costs
| Kind | Build | Running |
|---|---|---|
| Off-the-shelf widget, self-configured | £0 – £500 | Subscription, often per conversation |
| Grounded in your content, custom | £2,000 – £6,000 | Tens of pounds a month in usage |
| Connected to orders, stock or accounts | £6,000 – £15,000 | Usage plus any platform fees |
| Cybrial | Priced on application, from the hours | Your own accounts |
Source: UK market ranges, September 2026.
Note the first row. For a lot of businesses a configured off-the-shelf widget is genuinely the right answer, and a custom build would be money wasted. The custom case starts when it needs to know something only your systems know — stock, orders, account status — because that is where a widget cannot reach.


