Is an AI chatbot on your website worth it: the numbers, examples and a test
An honest calculation: how much an AI chatbot costs, which questions it can handle, which businesses it makes no sense for, and how to check before launch whether it will pay off.
Ads for AI chatbots promise more inquiries and less work. Sometimes that is true, and sometimes it is an expense that changes nothing. In this guide I show how to work it out with your own numbers before you pay for a deployment.
Here you will find the costs, the types of questions a chatbot handles well, the businesses I don’t recommend it for, a simple return formula and a way to test the idea before launch.
How much an AI chatbot really costs
In my offer, an AI chatbot for your website costs from PLN 2,490 net as a one-time fee and from PLN 199 net per month. In the first year that is at least PLN 2,490 + 12 × PLN 199, or PLN 4,878 net. In the following years only the subscription remains, about PLN 2,388 net per year.
These are “from” prices, because the final amount depends on the scope: the amount of knowledge, integration with email or a CRM, and the number of languages. Add the time you will spend preparing answers and reviewing conversations, especially in the first weeks.
Remember the legal obligations too. Since 2 August 2026 a chatbot must tell people that they are talking to AI (Article 50 of the AI Act). Adapting a chat is the AI Act package from PLN 490 net. I write more about the rules in the guide AI chatbots and the AI Act.
Which questions a chatbot handles well
A chatbot is good where the answer can be found in your offer or documents. It works best for:
- questions about the offer and the scope of services (“do you also install fire doors?”),
- questions about deadlines, service area and hours,
- preliminary questions about prices and how quotes are worked out, when you give ranges or rules,
- questions outside working hours, when nobody answers the phone,
- collecting contact details: a name, phone number and description of the matter reach you by email or in a CRM,
- questions about order status or procedures (returns, withdrawal, delivery).
These repetitive questions take the most time, and they rarely require your expert knowledge.
What a chatbot won’t do for you
A chatbot is not an expert in your industry and does not make decisions. I don’t recommend entrusting it with:
- binding quotes for unusual jobs,
- advice that carries responsibility (medical, legal, technical),
- conversations with an upset customer who needs a person,
- information that isn’t in your materials: then it should say “I don’t know” and pass the conversation on.
A well-configured assistant knows its limits. I described this in the guide A safe AI chatbot on your website.
When a chatbot pays off and when it doesn’t
The table below is my general assessment, not the result of a study.
| Situation | Does it pay off? |
|---|---|
| Lots of repetitive questions in emails, forms and phone calls | Usually yes |
| Customers write in the evening and at weekends, and you only reply in the morning | Usually yes |
| A broad offer that is hard to navigate (e.g. many products or types of services) | Usually yes |
| A few inquiries a month, each handled personally | Probably not |
| A website with one simple service and a phone number front and center | Probably not |
| The offer changes daily and nobody updates the knowledge | No, until you sort that out |
| Services where every case needs an individual expert assessment | Only for collecting contact details |
In the industries I know from my own deployments, the chat works as help with the offer and for collecting inquiries, not as a replacement for an advisor. That is the case, among others, on the websites Łazienki Giżycko and Drzwi Giżycko.
How to calculate the return: a simple formula
You don’t need a spreadsheet with dozens of columns. Two calculations are enough: time saved and extra jobs. The numbers below are illustrative assumptions; plug in your own.
Calculation 1: time saved
- Count how many repetitive questions you get per month (e.g. 60).
- Estimate how many of them the chatbot will handle without you (e.g. half, which is 30).
- Multiply by the time of one reply (e.g. 10 minutes): 30 × 10 min = 5 hours.
- Multiply by the value of your hour (e.g. PLN 100): PLN 500 per month.
Calculation 2: extra jobs
- Count how many inquiries per month are lost because a customer wrote in the evening and didn’t wait for a reply (e.g. 4).
- Estimate how many of them the chatbot will save (e.g. 1).
- Multiply by the average profit per job (e.g. PLN 800): PLN 800 per month.
The return
In this example the benefit is about PLN 1,300 per month, and the subscription costs PLN 199. A deployment for PLN 2,490 then pays for itself in about two to three months. If, however, you realistically save one job per quarter, the calculation looks completely different. So count carefully and enter numbers you know from your own emails and phone calls.
How to test the idea before deployment
Before you pay, check whether you have the material for a chatbot at all:
- Write down customer questions for two weeks. A simple list in a notebook is enough: the question, the channel, the response time.
- Mark the repetitive ones. If they make up less than a third, a chatbot will have little effect.
- Prepare the answers. If you can’t write them in a few sentences, the chatbot won’t answer well either.
- Measure when the questions come in. Many in the evening and at weekends is a good sign.
- Check your website. If customers can’t find basic information, sometimes it is enough to fix the website. You can do this with the free website test.
After deployment, measure the effect: the number of conversations, the number of contacts left and how many conversations a person from the company took over. It is worth seeing inquiry data in analytics, which I write about under the analytics and tags service.
What to watch out for after deployment
A chatbot isn’t an appliance you switch on and forget. In the first weeks, read the conversations and correct the answers, because that is when the questions nobody anticipated come up.
Also take care of three things: a notice that the customer is talking to AI, a privacy policy with a clause on processing conversations, and a set-up route to a person. If your customers mostly write on Messenger or WhatsApp, consider social media bots instead of a chat on your website.
Summary
An AI chatbot pays off when you get a lot of repetitive questions, customers write outside working hours, and you can describe the answers. It doesn’t pay off when there are few inquiries or each one needs an individual assessment. Before you decide, work out both calculations from this guide with your own numbers.
If you’d like me to check it with you, write to me. I reply within 24 hours, and new customers get 20% off their first project. You can also start with the free website audit, which will show whether the problem lies in the website or in the lack of an assistant.
Frequently asked questions
It depends on the number of valuable inquiries and how many of them the chatbot actually saves or captures. With the costs from my pricing, a few extra jobs a year or a dozen or so hours of work a month are usually enough to close the calculation. Work it out with your own numbers using the formula from this guide.
In a small business, usually not, and that is not why I deploy it. It answers repetitive questions, collects contact details and hands the conversation to a person when a case is unusual. It relieves people of phone calls with basic questions, but decisions and quotes stay on your side.
That is why the chatbot should answer only on the basis of your offer and documents, and refer to a person on matters of prices and deadlines. Before launch I test it on real customer questions, and after launch I read the conversations and correct the answers. I describe the safety rules in a separate guide.
Yes, you have to remember the GDPR and telling customers that they are talking to AI. Since 2 August 2026 this follows from Article 50 of the AI Act. I described the details in the guide on chatbots and the AI Act.
Usually not. If you get a few inquiries a month and handle each one personally, it is better to invest in a clear website, a Google Business Profile and a form. A chatbot makes sense when there are many questions, they repeat, or they arrive outside working hours.
Related services
Have a similar project?
Describe what you need and I’ll send a quote within 24 hours. New clients get 20% off their first project.
