Communicating with AI agents

What to hand to an agent and what to keep, how to brief one, reviewing a draft before it goes out, prompt injection, writing for readers who have AI too, and saying when AI was involved.

What to hand to an agent, and what to keep

AI is excellent at some messages and quietly bad at others; decide which is which before you hand anything over.

Briefing an agent well

An agent writes like a brilliant new hire who knows nothing about your situation; give it context, constraints and examples.

Review before it goes out

Your name is on the message, so check facts, tone and commitments before anything AI-drafted is sent.

Prompt injection: a message can talk to your assistant

Anything your agent reads can carry hidden instructions; treat incoming content as information, never as orders.

Writing for a reader who has an AI too

Your reader is busy and may be summarising you with AI, so write short, structured, checkable messages, and never send workslop.

Saying when AI was involved

Being open about AI help protects trust, and in the EU it is becoming a legal duty for AI systems that talk to people.