# What should an AI agent remember about my customers?

Only what makes the next task better, and what the customer would expect it to keep. A usual order, a preferred supplier, a delivery note. Never card numbers, passwords or health details. Let people see and delete what it holds. Memory that surprises a customer feels like surveillance, even when it helps.

By Surinder Thakur, AI Product Designer and Builder. Published 17 Nov 2026. Topic: [Trusting AI with money](https://www.surinder.design/work?type=trusting-ai-with-money).

## Why remember anything at all?

Because repeat questions waste everyone's time. An agent that knows the usual order can confirm it in 1 message.

That is the difference between a tool people tolerate and one they prefer.

## What is the test for each fact?

Ask 2 questions. Does it make the next task faster or more accurate? Would the customer expect it to be kept?

If either answer is no, the agent does not store it.

## What should it never keep?

Card and bank numbers, passwords, health details, and anything said in passing that the job does not need.

Those belong in the systems built to protect them, if anywhere. The agent asks for them when needed, then lets go.

## How do customers stay in control?

Show what the agent remembers, in plain words, and let people correct or delete it.

Memory people can see feels like service. Memory they discover feels like being watched.

## In 1 picture

1 test for every fact an agent could remember.

Does it help the next task?

- Yes (and the customer expects it): Usual order, Preferred supplier, Delivery note. Remember it.
- No (or it is sensitive): Card numbers, Passwords, Health details. Never store it.

## Next questions

- [How do you stop an AI agent from being tricked by a hidden instruction?](https://www.surinder.design/work/how-to-stop-prompt-injection-in-an-ai-agent): You cannot fully stop it, so you limit the damage.
- [What does the EU AI Act mean for a company using AI agents?](https://www.surinder.design/work/what-the-eu-ai-act-means-for-ai-agents): An agent that talks to people in the EU must tell them it is an AI.
- [How do you get customers to use a new AI agent?](https://www.surinder.design/work/how-to-onboard-customers-to-an-ai-agent): Say it is an AI in the first message. Show what it can do in 1 line, with an example.
- [How do you fix drop-off in fintech onboarding and KYC?](https://www.surinder.design/work/fix-fintech-onboarding-and-kyc-drop-off): People quit KYC when a step asks for trust it has not earned.

## Questions that lead here

- [Can an AI agent handle customer support without annoying customers?](https://www.surinder.design/work/can-ai-agents-handle-customer-support): Yes, with 3 rules. It answers only from your own documents and shows where each answer came from.
- [What data do I need before building an AI agent?](https://www.surinder.design/work/what-data-do-you-need-to-build-an-ai-agent): 3 kinds. The live records the agent reads, like stock, prices or open invoices.

Page: https://www.surinder.design/work/what-should-an-ai-agent-remember
