Skip to main content
VLADISLAV DEF / KONTURContact

Working on your task / Custom chatbot

A chatbot built around your business conversations

A chatbot makes sense when conversations contain a clear, recurring task: asking a few questions, collecting details, finding an answer or handing an inquiry to a staff member. If you are considering one, we will start with the conversation, channel and person's next step. We can then discuss feasibility and scope.

The conversation comes before the bot

Start with a few typical inquiries without personal data. What do customers ask? What information does a staff member need? Where does a conversation become an order, and where is clarification required? This forms the basis of a flow, rather than a list of fashionable features.

A bot can collect an inquiry, show options, ask required questions and pass the answers to a person. Telegram and website flows differ, so we choose the channel together with the task. Connecting your CRM and other systems is assessed separately.

  • An initial inquiry: service, city, starting parameters and a convenient way to continue the conversation.
  • Answers based on prepared materials: terms, service scope and steps to follow.
  • Event notifications, provided the source system offers an available way to receive them.
  • A handoff to a staff member when the question falls outside agreed boundaries or the customer asks for a person.

Does the bot need AI?

Choosing among a few services and collecting fields in sequence often needs only a standard button-based flow. It is easier to describe and test because each step leads to a known outcome.

AI can help when people ask questions in free text and answers need to be found in your materials. We then separately define the sources, permitted answers and situations requiring a handoff. The bot must not invent terms, prices or stock availability absent from the available data.

Not every business needs an AI bot. If there are few conversations and each requires a nonstandard decision, direct communication may be more convenient. Sometimes improving a page of terms or a short form is enough.

Where automation ends

We agree on who is responsible for new inquiries and how to see that a request has been picked up. A chat notification does not replace record-keeping: it can be missed, and several staff members may reply to the same customer.

The bot needs to say when it does not know an answer and explain the next step. If an integration is temporarily down, it must not confidently confirm a booking that does not exist. Failure behavior is part of the development task.

What we check before launch

We test with sample data. One successful reply does not establish that the entire chain is ready. Without access to a required system, connecting it remains an open question rather than a promised feature.

  • A normal inquiry travels from the first message to the required record or staff handoff.
  • Missing fields, unexpected answers, repeated submissions and requests for a person are handled clearly.
  • The bot uses agreed materials and does not invent missing information.
  • External service errors are visible to the responsible person, and a retry does not create an extra order.
  • The client has access to the result, a description of the flow and a clear process for updating materials.

What determines development scope

The estimate covers conversation branches, channels, answer sources, integrations and data storage requirements. External service usage costs and ongoing support are discussed separately; inventing a price for your task in advance would be misleading.

Tell me who will talk to the bot, give a few typical questions and specify where the result should go. If there is no flow yet, we will begin by describing one complete conversation. You do not need bot tokens or passwords for the first message.

Next step

Let’s discuss your task

Describe one conversation you would like a bot to handle: who writes, what they ask and who should receive the result. We can discuss the flow and whether it is feasible.

A message to get you started

Hi Vladislav! I read your ‘Business chatbot’ page. I need a bot in [channel] for [task]. A typical customer question is [example]. The result should go to [where/who]. What do you need for an estimate?

Adjust it to your task. Opening a chat does not send a message.