PricingHow much does an AI chatbot cost?

Almost all of the variation between one quote and another comes down to a handful of things. If you know where you sit on each, you know roughly what kind of project you have.

The first is a subscription tool you configure yourself. You pay monthly, often by conversation or by seat, and you do the setup. The second is a system built for your business, quoted as a project, usually with a monthly amount for hosting and upkeep.

We build the second kind, and we will still tell you plainly: if what you need is a widget answering a dozen common questions on a simple website, a subscription tool may well be all you need. Start there. Come back when it starts costing you customers because it cannot see your calendar or your order data.

The moment a business genuinely needs the second kind is usually the moment the answer has to be true rather than generic. Looking up this customer's order. Offering the times you are actually free. Applying the policy you actually wrote.

A build price is not the whole number, and a quote that only shows the build price is not being straight with you.

Quotes in this market are hard to compare because they rarely describe the same thing. These questions make them comparable, and they work on us as well as on anyone else.

Send the same brief to three vendors and the quotes will not be close. That is not always someone trying it on. Usually they have each quoted a different job, because the brief left room to.

The first has priced a widget that answers from your existing text. The second has priced that plus a connection to your calendar. The third has priced the version that survives a wrong answer in front of a customer, with testing, guardrails and a handover path. All three are honest. Only one of them is what you meant.

The way to make the quotes comparable is to fix the parts that move: how many systems it touches, how many channels it answers on, who writes the knowledge, and what happens when it does not know. Once those are written down, the numbers converge and the differences that remain are real.

Take a dental practice with one location. They want the website to answer opening hours, insurance questions and parking, and they want it to book into their real calendar rather than take a message.

The answering half is straightforward: the questions repeat, the answers are already written down somewhere, and none of it is sensitive. On its own, that is the cheapest kind of project there is.

The booking half is what changes the price. Now something has to read live availability, hold a slot, write the appointment back, and behave correctly when two people book the same time. That is a build, not a subscription tool.

The same practice asking for the same thing across a phone line, a website and WhatsApp, in two languages, with a rule that anything clinical goes straight to a human, is a different project again, and it moves up a row.

If your budget is genuinely tight, do not start with the build. Start with a subscription tool, point it at your existing pages, and watch what people actually ask it for a month.

That month is worth more than any quote. You find out which questions repeat, which ones the tool cannot answer, and whether the thing costing you customers is information or action. Most owners are surprised: they assumed they needed booking, and the log says four in five questions were about hours and price.

If the log says information, keep the tool and spend nothing more. If the log says action, you now have a specification written by your own customers, which is the cheapest brief anyone will ever hand a vendor, and it will take a chunk out of the build cost because nobody has to guess.