Chatbots: Back to Basics
A practical introduction to chat and knowledge services: useful answers, approved information, clear next steps and a reliable handover.

Reviewed and updated 22 October 2026. Original publication date retained.
Begin with the conversation you want to support
A chat service should help someone complete a useful task. That might mean finding approved guidance, understanding a service or being directed to the right team. Start by identifying the audience and the questions the service should handle. A clear purpose makes it easier to choose the right mix of guided choices, search and generated responses.
Choose the right approach to answers
Some answers can be provided through a defined flow. Others may use retrieval from approved sources or an AI model to help interpret the question and prepare a response. The design should reflect the information available and the consequences of an incorrect answer. There should be a clear response when the material does not support an answer, rather than an expectation that the system must always provide one.
Make information ownership part of the service
A knowledge service needs current material and someone responsible for maintaining it. Review permissions, conflicting documents and the route for correcting an answer. When personal or sensitive information is involved, include appropriate specialist input. The ICO AI resources are a starting point for researching relevant data-protection questions, not a substitute for reviewing the particular service.
Design a clear handover
Customers and staff need to know what happens when a question needs human help. Agree the destination, the information passed on and what to do outside normal service hours. A service can be available when the team is offline without implying every task can be completed immediately. Make the next step clear and test it alongside the normal conversation.
Test the service people will actually use
Use real-world phrasing, incomplete questions, out-of-scope requests and examples from the people who know the work. Check accessibility, mobile use and whether the information leads to a useful outcome. Explore chat and knowledge, integration and software and AI Governance for the design and operational support behind the conversation.







