Insights · AI at workFive rules before your team puts client data into AI

1. Use business accounts, not personal ones

Free personal accounts are where most data leaks start, because nobody has checked their settings. Put your team on business plans of the tools you choose, and check each one's data controls, including whether your content can be used to train models. Turn off anything you don't need.

2. Decide what never goes in

Write a short "never" list and keep it visible. A typical one: Social Security and ID numbers, bank and card numbers, passwords and logins, health information, and anything under an NDA or confidentiality clause. When in doubt, remove names and numbers before pasting.

3. A person signs off on anything that leaves the building

AI drafts; people send. Anything that reaches a client, a vendor, a lender, or a bank account gets reviewed by a named person first. This one rule prevents most embarrassing mistakes, and it's what makes the rest of the team comfortable using the tools.

4. Ground answers in your own information

AI is most reliable when it works from your approved documents: price lists, contracts, past proposals, and procedures. Keep a small, current library of those and point the tools to it, rather than letting them guess.

5. Write it down and train on it

Rules nobody has read don't work. Put these on one page, walk the team through it with real examples from their own work, and revisit it after a month. Ask what's confusing and what's slowing people down, then adjust.

Want help putting this in place?

The one-page policy is part of the first two weeks of my AI enablement program, along with training your team on the tasks where AI saves them the most time.