The job
What exact task will the agent do, and how will you know it did it well?
Free PDF
The AI Agent Safety Checklist
The questions to answer before you connect an AI agent to your email, CRM, files or payments. 24 questions in 8 short sections, in plain English.
- What the agent can see, and whether it needs all of it
- What it can send, change or delete, and what needs your approval
- Where your data goes and whether it trains someone's model
- How to spot the tricks, and how to turn it off fast
Where should I send it?
What's inside
What it can see
Which accounts, folders and records will it be able to read?
What it can do
Can it send, change or delete anything, or only read and draft?
Where your data goes
Which companies receive your data: the agent vendor, the AI model provider, anyone else?
Tricks and mistakes
What happens if an email or document tells the agent to do something you didn't ask (prompt injection)?
Control
Whose login does it use, and is that account protected with multi-factor authentication?
Your business, your name
Does it tell people it's AI when it should?
Worth it
What does it cost per month at your real volume, including usage fees?
Each section has three questions. The first one is shown above; the PDF has all 24, with space to write your answers.
After you read it
If one question stops you, that's the one to check.
Book a free 30-minute call and we'll go through it together.