A chatbot for your business: when it pays for itself and when it just annoys clients
People talk about bots in two ways: "this is the future" and "this is the irritating box that jumps out after three seconds". Both are true — the question is which job the bot is doing.
Below: where it genuinely saves money, where it gets in the way, and what it costs.
Want the routine automated?
Python bots for enquiries, bookings and auto-posting listings to Bazoš.
What a bot does well
One thing only, but an important one: repetitive actions with no variation.
Taking enquiries. Someone sends a request and you are notified immediately — not when somebody next opens the inbox. In practice that is often the difference between "answered in twenty minutes" and "answered the next day", by which point the client has written to somebody else.
Booking. The client picks a time, staff get a message, one tap confirms it and the slot closes on the site automatically. That is exactly how booking works at Krása štúdio OK: no spreadsheets, no double bookings.
Answering standard questions. Opening hours, address, prices, how to get there, what to bring. That is 60–70 % of incoming messages in almost any business.
Notifications inside the team. A new enquiry, a new order, an error on the site — into the chat that is open anyway.
Routine with no client involved at all. Auto-posting listings on Bazoš, for instance, for anyone who posts there regularly: by hand it eats hours a week.
What a bot does badly
Honestly, point by point:
Selling something complex. If the decision needs a conversation, a bot only gets in the way.
Handling an unhappy client. Someone with a problem is infuriated by a bot. This needs a person, quickly.
Answering what was never built into it. A bot that replies "I did not understand" to every non-standard question works against you.
Substituting for a process that does not exist. If enquiries get lost because there is nobody to handle them, a bot merely speeds up the loss.
Telegram or a widget on the site
That is usually the choice, and it depends on where your client is.
| Site widget | Telegram | |
|---|---|---|
| Who it catches | someone already on the site | someone arriving from anywhere |
| After the tab closes | conversation lost | conversation stays |
| For your staff | a separate panel | the phone already in their hand |
| Second touch | none | you can write later |
| Speed of reply | requires sitting in the panel | arrives as an ordinary message |
The practical conclusion: a widget fits a high-traffic site where somebody is on duty. In every other case Telegram is more convenient — for the client and for you.
The reason is simple and not technical: any system that requires staff to open a separate panel does not take root in a small business. Between two treatments in a salon there is one minute, not ten.
What it costs
A bot for a specific job — from 500 €. Changes at 25 € an hour.
The price depends not on the number of messages or the "sophistication of the AI" but on integrations:
- a standalone bot answering from a script — the lower bound;
- a bot linked to the site (enquiries, booking) — the middle;
- a bot linked to a calendar, stock, payments or an accounting system — the top.
For comparison: an employee answering the same questions by hand for an hour a day costs more than that in the first month.
What to measure to know whether it paid
Three numbers worth recording before launch and checking a month later:
- How much time standard replies take now. Roughly is fine.
- How long passes between an enquiry and the first reply.
- How many enquiries are lost — unanswered, forgotten, buried in email.
A bot pays for itself when the second and third numbers change. The first is a bonus.
Do you know how many enquiries the site actually brought?
I set up GA4, Tag Manager and conversion tracking so the numbers match reality.
Mistakes that turn a bot against you
Popping up three seconds after arrival. The person has not yet worked out where they are.
No route to a human. A "talk to a person" button belongs in the very first menu.
Silent at night. A bot that says nothing outside working hours is useless at exactly the time it could be working instead of you. "We will reply in the morning, meanwhile here are answers to common questions" beats silence.
Too many steps. If there are five questions before the enquiry, not everyone reaches the end.
Not reporting the enquiry anywhere. The classic: the bot collects the data, it lands in a database, and nobody looks at it.
When a bot is not needed
Plainly, because this is more common than it seems:
- Few enquiries. Five a week is nothing to automate.
- Every enquiry is unique. Project work, consulting, complex services.
- The problem is not reply speed but that nobody finds you. Then the money belongs in the site and its visibility, not a bot — start with checking the site.
What building one looks like
- A half-hour conversation. Which questions come up most and what eats time.
- A script. What the bot can do and where it hands over to a person.
- Development — a few days to two weeks, depending on integrations.
- Testing on real enquiries.
- Refinement based on what people actually ask.
The fifth step matters more than the first four: real questions always differ from the expected ones.
If you need a bot
I write bots in Python — enquiries, booking, notifications, auto-posting. From 500 €, usually one to two weeks. Details on the bot development service page.
I have lived in Slovakia for over ten years, work as a Slovak s.r.o. and issue a faktúra with an IČO.
Related: the salon case with Telegram booking and how Slovaks choose a contractor.
Shall we talk about your project?
Send a few lines about what you need. I reply within 24 hours, no obligation.
