Getting a chatbot built — what it costs, what works and when it pays off
We've all "talked" to a chatbot that answered nothing. You type a simple question, and the bot cheerfully sends you back to a menu that's missing exactly the option you need. That kind of chatbot has given the whole category a bad name. But a lot has changed in the past few years: a chatbot trained on your own content can in fact give a normal answer. This piece explains what getting a chatbot built really means today, what it costs and when it's worth it.
Because a chatbot isn't a goal in itself. It's a tool that does one thing well: automatically answering the same questions your customers keep asking — day and night, in several languages, without you having to pick up the phone. The only question is whether your business gets enough of those questions to make it worthwhile.
Three kinds of chatbot (and why that difference matters)
Menu bot
The old-fashioned variety: buttons and fixed paths. Cheap, but frustrating the moment the customer asks something outside the script. Fine for very simple things, poor for the rest.
AI bot on your own content
An assistant trained on your documentation, FAQ and web copy. Answers in normal language and understands questions you didn't literally foresee. This is usually what people now mean.
General AI without guardrails
A bot trained on the whole internet but knowing nothing specific about your business. Sounds clever, but without guardrails it'll happily invent your opening hours. Don't deploy without reining it in.
The difference that counts: a good business chatbot is trained on your information and stays within it. It invents nothing, it refers to what you've written down, and when it doesn't know something it says so honestly and hands over. That last part matters more than it sounds — a bot that says "I don't know, let me connect you" is infinitely better than one that invents something wrong with a straight face.
Rule of thumb: a chatbot is only as good as the content it's trained on. If you've nowhere written down neatly how you work, the bot has nothing to draw on either. Often "getting our FAQ in order" is the first, most valuable step — with or without a bot.
When does a chatbot pay off?
A chatbot earns its keep on repetition. If you keep getting the same questions, a bot takes that work over and you keep time for the questions that really matter. A few situations where it usually pays off:
- You get the same questions daily: opening hours, prices, availability, "do you also deliver to…".
- Your customers are in different time zones or ask things in the evening — a bot never sleeps.
- You serve customers in several languages, for example Dutch and German in the border region.
- You lose enquiries because people don't want to wait for an email reply.
And when not? If you get few repeat questions, if every customer question is unique and complex, or if your audience specifically expects personal contact and would find a bot distant. A chatbot for a handful of questions a month is an expensive solution to a problem you'd also solve with a good FAQ page.
What does getting a chatbot built cost?
There are two cost items here, and it's important to keep them apart. The first is the build: setting up, training on your content, configuring and testing. The second is running it: an AI chatbot uses computing power, and that costs a small amount per conversation. The more people talk to it, the higher those running costs — usually still modest, but not a free lunch.
A simple menu bot is the cheapest to build and run, but also delivers the least. An AI bot on your own content costs more to set up, but gives better answers and saves more time. Which makes sense for you depends on how many questions you get and how complex they are. I'd rather work that out honestly with you than push the most expensive option "because AI".
Tip: start small and focused. Train the bot on your five to ten most-asked questions, put it live, and watch for a few weeks what people actually ask. Often you discover questions you hadn't thought of — and you make those properly afterwards. A bot that gets better in steps is more reliable than one that has to do everything in one go.
The pitfalls you really want to avoid
A chatbot can also work against you. The biggest risks:
Inventing
AI without guardrails invents answers. For a business that's dangerous — wrong prices or promises are worse than no answer. Good setup keeps the bot inside your information.
No escape
A bot without a clear "talk to a human" exit frustrates. There must always be a neat hand-over when the bot can't work it out.
Privacy
Customers sometimes type sensitive things into a chat. A GDPR-aware setup and clarity on what happens to data isn't a detail but a requirement.
Set-and-forget
A bot you put live and never look at again ages. Glancing now and then at what people ask keeps it sharp and useful.
Multilingual, across the border too
One of the nicest properties of an AI chatbot is that multilingualism comes along almost for free. The same bot can serve a Dutch customer in Dutch and a German one in German, without you having to write everything down twice. For business owners in the border region with Germany that's a concrete advantage: you're reachable in both languages in the evening and at weekends, even when you're not on the phone.
Chatbot or something else after all?
Sometimes the most honest conclusion of a chatbot conversation is that you don't need a chatbot at all. That sounds odd coming from someone who builds them, but it's true. A few alternatives that often solve the same problem more cheaply and simply:
- A good FAQ page. If you mainly get the same five questions, a neat FAQ answers them just as well — without monthly costs. Often it's the first step anyway, because it's also the basis a bot is trained on.
- A booking calendar. Do most questions revolve around availability and appointments? Then a direct booking calendar takes away more friction than a bot talking about the same thing.
- A smarter contact form. Sometimes the customer just wants to pass something on, and a short, clear form is nicer than a conversation with a bot.
I'd rather first check whether one of these solves it before we build a bot. A chatbot is fun toy, but it has to deliver something — otherwise it's an expensive widget blinking in the corner of your screen without anyone being better off.
Frequently asked questions
Does a chatbot replace my customer contact?
No, and that's not the goal. A good bot catches the repeat questions so you keep time for the conversations that really matter. The personal contact stays — only the boring questions disappear.
Can the bot say something wrong?
If it's well set up and trained on your content, it stays within what you've written down and hands over when it doesn't know something. The risk of nonsense lies mainly with unguarded AI — exactly what you want to avoid.
How quickly is a chatbot live?
A simple bot on your most important questions can be running in a few weeks. After that it's a matter of refining, based on what people actually ask.
In short
Getting a chatbot built is neither a hype nor a magic cure. It's a level-headed trade-off: do you get enough repeat questions to hand the work off, and is your content good enough to give the bot something to say? If yes, a well-set-up AI bot saves you time and keeps you reachable outside office hours — in several languages. If no, I'll say so and we look at something simpler.
Curious whether it pays off for you? Book a no-obligation call or first read the more in-depth piece on an AI chatbot for your business.
Related articles
AI chatbot for your business: what works, what doesn't?
A deeper look at when an AI assistant on your own content adds value.
What belongs on a good business website?
The basics that build trust — a good FAQ is part of it.