Telegram and WhatsApp bots, AI assistant for business

A business chatbot is a bot in WhatsApp, Telegram or on your website that answers common questions around the clock, takes enquiries and bookings, and passes to a manager only what needs a human. It suits businesses that get the same questions every day: service companies, clinics, schools, shops. Pricing starts from €500, with a timeline from 14 days.

Who needs a bot and what it handles

• The same questions every day: prices, opening hours, address, delivery, “is it in stock?”. The bot answers instantly and your team stays focused.
• Messages at night and on weekends go unanswered, and the customer goes to whoever replied first.
• Enquiries from messengers get lost in managers’ personal chats.
• Bookings by chat take hours — the bot offers free slots and records the booking.
• Lead qualification: the bot asks a few clarifying questions, so the manager receives a clear request.

More on what slow replies cost you is on our page about slow responses to customers, and on lost enquiries — leads get lost.

Scripted bot or AI assistant

Criterion	Scripted bot (buttons and menus)	AI assistant
How it answers	Through predefined flows	In its own words, based on your materials
Best for	Few request types, a clear process	Many different questions, a large price list or knowledge base
Predictability	Full: it only says what is scripted	High, if sources are restricted and there is a human hand-off
Maintenance	Update the flows when things change	Update the materials it answers from

A mix often works best: buttons for enquiries and bookings, AI for open questions. Read more in AI agents for business and on our AI services page.

Stages and timeline

1. Collecting typical requests. We go through real conversations and the most frequent questions, and decide what the bot handles alone and what it passes to a person.
2. Flows and copy. We write the dialogues in your tone of voice and agree them with you.
3. Build and integrations. The bot, leads into your CRM or a spreadsheet, team notifications and, if needed, calendar bookings.
4. Testing. We run real questions through it and check that it does not make things up and calls a human in time.
5. Launch and tuning. In the first weeks we review real conversations and refine the flows.

A bot with typical flows takes from 14 days. An AI assistant on a large knowledge base or several channels takes longer.

What affects the price

Factor	Effect on price and timeline
Base bot: flows, lead capture, manager hand-off	from €500, from 14 days
AI answers based on your materials	Preparing the knowledge base and testing answers
Number of channels (WhatsApp, Telegram, website)	Each channel has its own limits and setup
CRM, calendar or payment integration	Each integration adds work
Several languages	Translated flows and checked answers

Third-party costs may come on top of our price: WhatsApp Business Platform messaging fees or AI model usage. We name them up front.

Tools and platforms

WhatsApp Business Platform, Telegram Bot API, website chat widgets; AI models restricted to your own materials; API integrations with CRMs, Google Sheets and calendars. We write bots in Node.js/TypeScript or Python and deploy them on your server or in the cloud. If you only need appointments, compare with our online booking service.

An example from our work

For a university admissions website we built an AI assistant widget that answers applicants’ questions about admission and programmes, while form enquiries go straight into the learning platform. The admissions office gets fewer repeated questions. Details are in the university admissions case study. In a Shopify supplement store, orders and enquiries arrive automatically in the team’s Telegram and the CRM. Our own website also sends every enquiry to Telegram.

What to prepare before we start

• 20–30 real customer questions from recent conversations — they are the backbone of the flows.
• Current prices, terms and hours, plus the answers your team gives to the most common objections.
• Hand-off rules: which questions the bot never handles alone — complaints, refunds, individual discounts.
• Where leads go: CRM, spreadsheet or team chat, and who is responsible for them in and out of hours.

What a bot should not do

A good bot does not pretend to be a person, does not promise anything outside your terms and does not trap customers in endless menus. Every step needs a “talk to a manager” option. An AI assistant does not invent prices or deadlines: if the answer is not in your materials, it says so and passes the conversation on. And a bot does not replace a fast team response — it buys the team time by recording the request, collecting details and giving a realistic reply time.

What to measure after launch

• Share of conversations closed without a manager — how much routine the bot really took over.
• Out-of-hours enquiries — the ones that used to get lost.
• Time to first reply and time to a manager’s reply after hand-off.
• Questions the bot could not answer — a ready-made list for updating your materials.
• Where customers drop out — a sign that a flow is too long or unclear.

How to choose a contractor

• Is there a human hand-off, and can the manager see the conversation history?
• Where does the AI get its answers, and what does it do when it does not know?
• Who owns the bot, tokens and data — they should be in your name.
• Who updates the flows when prices or services change?
• Are third-party costs included for messengers and AI models?

If messenger enquiries are already slipping away, start with where lost leads go and how to stop it: a bot closes only some of those gaps, and rules and a CRM close the rest.

Want a bot that handles routine questions and never loses a lead? Tell us what customers ask most often and we will suggest a flow and a price.