Before putting customer data into AI
What to remove and check before sharing customer information, including identifying notes, combinations of details and retention settings.
Pasting a whole table into AI is convenient, but customer information needs a closer look first.
Start here
Changing a name may not be enough. A date, location and order history together can still identify someone.
- Practice with made-up data.
- For real work, share only the fields needed for the task.
Before pasting
- Are names, phone numbers, addresses, emails or order IDs necessary? Remove them if they are not.
- Check free-text notes and requests. A note such as “the manager in the third-floor office” can identify someone.
- Never include bank account numbers, card numbers or national identification numbers.
- Check whether you are using a work account or a personal account.
- Check the service’s training settings and data retention period.
What to remove: a fictional example
| Original field | What to share |
|---|---|
| Name, such as J. *** | Remove |
| Partially masked phone number | Remove |
| Order ID | Remove |
| Request: “For the manager on the third floor” | Remove |
| Order date, item and amount | Keep only if essential, and check whether the combination could identify a person |
Settings and obligations
- Turning off training controls whether the provider uses conversations for training. It does not mean nothing is retained, or give you permission to share customer information.
- Privacy obligations still apply. If you handle substantial customer data, check the rules and procedures that apply to your business.
If you have already shared it
- Stop adding more information.
- Delete the conversation. This does not guarantee that every copy held by the provider is erased.
- Check the provider’s retention and deletion guidance, along with your business’s customer-data procedures.
Need something like this?
If you have similar repetitive work, I can build it for you. Start by telling me which files you use and how you work today.