Chatbot

A chatbot is a program that talks to people by text. In support, it exists to solve what repeats — and, above all, to hand off well what cannot solve itself.

The two types, and why the difference matters

Rule-based (or flow) bot. Follows a fixed script: an options menu, conditions, canned replies. It is predictable, cheap and never makes up an answer. In exchange, it breaks when the customer writes something outside the script.

AI bot. Interprets what the person wrote in natural language, with no menu. It handles the unexpected and replies in context. In exchange, it needs clear boundaries — otherwise it answers what it should not.

In practice, good operations use both: rules for what is deterministic (a duplicate invoice, order status) and AI for what is open-ended.

Where a chatbot genuinely helps

• Outside business hours — that is where it stops competing with a human and starts replacing silence.
• The usual questions — price, timeline, address, payment methods.
• Initial qualification — collecting the three pieces of information the salesperson would have to ask anyway.
• Triage — routing each conversation to the right queue.

The mistake that makes customers hate it

The problem is almost never the bot itself — it is a bot with no way out. A flow that does not understand, repeats the question, and never offers "talk to an agent" turns automation into a wall. The practical rule: at any point in the flow, there must be a path to a human — and the human should receive the history along with it, so the customer does not have to start over.

See also humanized support, which is not the opposite of automation.