On this page
How a support bot works
A support bot sits between your customers and your team. When a question arrives, it first tries to answer from a knowledge base you control — delivery times, refund policy, how to reset a password, where to find an invoice. If it can answer, the customer gets a reply in seconds. If it cannot, or the customer asks for a person, it creates a ticket and a live agent takes over the same chat.
Answers can come from structured menus (“Orders”, “Payments”, “Technical help”), from keyword matching, or from an AI model that searches your documents and writes a reply grounded in them. We choose the approach based on how varied your customers’ questions are, and we can combine them.
Signs you need one
- Your team answers the same handful of questions many times a day.
- Customers message outside working hours and wait until morning for simple answers.
- Support happens on one person’s phone, with no record once the chat is closed.
- Customers get different answers depending on who replies.
- You cannot say how many queries you receive or how long they take to resolve.
This is common for D2C brands handling order queries, SaaS products with a Telegram community, clinics fielding appointment questions, and coaching institutes answering fee and schedule questions from students and parents.
What’s included
- Knowledge base that your team edits from an admin panel, so answers stay current.
- Order and account lookups — the bot can fetch a delivery status or subscription date from your system after verifying the customer.
- Ticket creation with category, priority and the full chat attached.
- Live agent handoff — agents reply from a team inbox or a staff group, and the customer sees the replies in the same chat.
- Business hours and away messages, with clear expectations about when a person will reply.
- Unanswered question log, showing what the bot could not handle — often the most useful report for improving your documentation.
- Satisfaction check after a conversation closes, if you want it.
AI answers or fixed answers?
| Approach | Strengths | Watch out for |
|---|---|---|
| Menus and fixed answers | Predictable, cheap to run, easy to approve | Feels rigid when questions vary a lot |
| AI with retrieval from your documents | Handles varied wording and follow-up questions | Needs guardrails, testing and a clear fallback to humans |
| Combined | Menus for common paths, AI for free-text questions | Slightly more setup |
Where we use AI, the model answers only from your own material and hands over to a person when it is unsure. Our AI Chatbot Development page explains that approach in more detail.
How we deliver it
- Collect a sample of real customer questions from chats and emails, and group them.
- Write or clean up answers for the most frequent groups with your team.
- Build the bot, the knowledge base editor and the agent handoff.
- Connect order or account lookups where needed, with proper verification.
- Test with real questions, launch, and review the unanswered log in the first weeks.
Measuring support quality
Once support runs through a bot, you get numbers you probably never had before. The admin panel shows how many conversations came in, how many the bot resolved on its own, how many went to a person, how long customers waited for a human reply and which topics come up most often.
- Topics that keep reaching agents are candidates for new knowledge base answers.
- Long waits at certain hours show where you need more staff cover.
- Repeated questions about one product or policy often point to a problem worth fixing at the source.
- Satisfaction responses, if enabled, show whether automated answers are actually helping.
We review these with you in the first weeks after launch, because that is when a small change to the knowledge base usually makes the biggest difference.
Effort, and mistakes to avoid
Effort depends on the size of the knowledge base, whether AI answers are involved, the number of systems the bot looks up, the number of agents and teams, and languages. The mistakes we see most often:
- Hiding the human. Customers who cannot reach a person get angry quickly. Make the handoff obvious.
- Stale answers. A policy changes and the bot keeps quoting the old one. Your team must be able to edit answers themselves.
- Sharing private data without checks. Order and account details should only be shown after verifying the customer.
- Launching without real questions. Build from what customers actually ask, not what you assume they ask.
Frequently asked questions
Can the same knowledge base serve WhatsApp and our website too?
Yes. We can keep the knowledge base in one place and connect it to a Telegram bot, a website chat widget and WhatsApp, so answers stay consistent across every channel.
Will the bot give wrong answers?
Menu and fixed-answer bots only say what you have written. AI-based bots are limited to your own documents and are built to hand over to a person when unsure, and we test them against real questions before launch. No system is perfect, which is why the human handoff matters.
How do agents reply to customers?
Agents reply from a shared team inbox or a private staff group. The customer sees the replies inside the same bot chat and does not need to contact anyone separately.
Can the bot check order status?
Yes, if your order system has an API or database we can read. The bot verifies the customer first, for example with an order number and phone number, before showing any details.
Can it work outside business hours?
Yes. The bot answers what it can at any hour and tells customers when a person will pick up anything it cannot resolve.
Talk to us about customer support bots
FAQ handling, ticket creation and clean handoff to a live agent without leaving the chat.