How do I teach an AI assistant what our business terms mean?

An assistant may know general plumbing language but not how your company uses it. A fictional plumbing company could give it this glossary before asking it to draft customer replies or summarise job notes.

Glossary for our plumbing company

Use these definitions when reading our job notes or writing for our team. If a term is unclear, ask rather than guess.

• Callback: A return visit to address a problem with work we recently completed, not a new job. Example: ‘Book a callback for the tap we repaired yesterday; it is still dripping.’
• Rough-in: The stage when pipes and connections are installed before walls are closed and fixtures are fitted. Example: ‘The bathroom rough-in is ready for inspection; the sink is not installed yet.’
• Service plan: Our ongoing maintenance agreement, not a one-off repair booking. Example: ‘This customer has a service plan; check their agreement before quoting for the visit.’

End of glossary

Now ask: ‘Draft a reply to a customer who needs a callback after yesterday's tap repair.’ The assistant should treat it as a return visit about existing work, rather than offer a new installation. Check the draft, especially any promises about charges or timing: the glossary does not tell it your pricing or availability.

Keep the glossary somewhere your team can maintain, then put a copy where the assistant can use it:

• ChatGPT: Put a short glossary in Custom Instructions, or add a glossary file and instructions to a Project.
• Claude: Add the glossary to a Project's knowledge files and tell it to use those definitions in the project instructions.
• Copilot: Where your version supports custom instructions or attached files, add the glossary there. Otherwise attach or paste it into the conversation with your request.

Features vary by plan and version. Replace old copies when your terms change, and avoid including customer details or other confidential information unless your organisation has approved that use.