Website chatbots failed. The people who installed them say why.
The people who installed them are the ones warning you off. Their reasons are good, and they point at what the replacement has to do differently.
If you ask a room of small business owners about website chatbots, you will not get a debate. You will get a groan.
And the strongest opinions come from the people who put them in.
One web person wrote that they now steer their small business clients away from chatbots completely. Another said a client got seven chats in six months.
They are right. This piece is not going to argue with them.
Why they failed
Four reasons, and they are all about design rather than technology.
They did not know anything. Most were a menu with three options and a form at the end. Ask them anything real and they collected your email. That is not answering. That is a slower contact form.
They were built to stop you reaching a person. Big companies used them to cut support costs. Customers worked that out fast, and the whole category got the blame.
They popped up uninvited. Three seconds after landing, covering the text you came to read, with a fake typing dot to look human. Everyone learned to close it without looking.
Nobody checked what they said. They got switched on and forgotten. A year later one was still quoting last year's prices and nobody had read a single transcript.
Sometimes nobody was going to ask anyway
The seven-chats-in-six-months case deserves its own mention, because it is the most common outcome and the least discussed.
If two hundred people visit your site a month and most already know what they want, there is very little for anything to answer. The tool was not bad. It was pointed at a problem that was not there.
Before anything else, work out whether you are in that group. Here is how to tell, and it is fine if the answer is no.
What to ask for instead
If you do have real questions arriving, the four failures above turn into four requirements. Any tool worth switching on should pass all of them.
It answers from what you wrote. Your prices, your area, your hours. Not a guess, and not a general answer about your industry.
It says when it does not know. And hands over quickly. Test this before you buy, with a question it cannot possibly answer.
It waits to be asked. No pop-up three seconds in. If someone wants it, they will find it.
You can read what it said. And fix a wrong answer in under a minute. If you cannot see the answers, you cannot trust them.
The thing that changed
Old chatbots could only match words. You had to guess every way someone might phrase a question, and you always missed some.
Newer ones understand that "do you come out to HG1" and "is Harrogate too far" are the same question. That is a real difference and it matters.
It also creates a new risk. A tool that understands anything can answer anything, including things you never told it. That is why the second requirement above is the important one.
Where we stand
We make one of these, so read the last section with that in mind.
We built hailera to answer only from what the owner wrote down, and to say so when it does not know. You check everything before a customer sees it.
That is a narrower promise than most of this industry makes. It is narrow on purpose. The wide promises are what got the category its reputation.
Written by the team at hailera — we help small businesses answer customer questions without being there.