Do and don't rules for your supermarket AI agent
This guide gives a supermarket plain do and don't rules for an AI agent answering shoppers, covering stock claims, allergens, offers, pickups and when to call a person.
Runs in your browser. Nothing you enter is sent anywhere or saved.
Before you start
An AI agent follows the rules you give it. In a supermarket, the most useful rules are short and concrete: what to say about stock, what to leave to staff, and how to talk to a shopper who is cross. Vague instructions give vague replies.
Copy the rules you agree with into your agent's instructions, and edit the rest to match your store. Each rule comes with a reason so you can explain it to your team and decide whether it fits the way you trade.
Tick each one as you go. 0 of 10 done.
[ ] Do give opening hours for the specific day the shopper asks about. [ ] Don't promise an item is in stock unless it was checked recently. [ ] Do say when stock information was last checked. [ ] Don't answer allergy or ingredient questions yourself. [ ] Do ask for the order name before discussing a pickup order. [ ] Don't quote a promotion that has ended. [ ] Do offer a sensible alternative when something is unavailable. [ ] Don't argue with a complaint or defend the store. [ ] Do hand over refunds, recalls and age-restricted items to staff. [ ] Don't message anyone who has not written to you first.
Turning rules into instructions
Write each rule as one plain sentence, in the voice you want the agent to use. For example: "If someone asks about allergies, say a team member will reply and pass the chat on." Short, direct lines work better than long paragraphs full of exceptions.
Add two or three sample replies in your own style. Agents copy the tone of examples closely, so a friendly local-shop voice in your samples will come through in the chats.
Keep the rules alive
Review the guide whenever the store changes.
- After any complaint caused by a wrong or cold reply.
- When you start a new service such as delivery or a pharmacy.
- Before big seasonal periods, when questions change.
- When new staff join and need to understand the hand-over rules.
FAQ
How many rules should I give my agent?
A dozen clear ones is plenty to start. Cover stock, hours, hand-overs and tone. You can add more after reading a few weeks of real chats and seeing what is missing.
Can the agent break these rules?
Agents can make mistakes, which is why weekly reading matters. Clear, specific rules reduce errors, but sensitive topics should always go to a person as a safeguard.
Should the rules be different for each store?
The core rules should match. Local details such as hours, car parks and counters should differ, so keep one shared rule set and a short local sheet for each branch.
More free tools for supermarkets
- First-reply checklist for supermarkets
- Hand-over rules checklist for supermarkets
- First-week checklist for supermarkets
- Weekly chat review checklist for supermarkets
- Busy-season checklist for supermarkets
- Customer privacy checklist for supermarkets
- Greeting messages for supermarkets
- After-hours replies for supermarkets
- Booking and order confirmations for supermarkets
- Reminder messages for supermarkets
- Apology and complaint replies for supermarkets
- "Let me check" replies for supermarkets
- Follow-up messages for supermarkets
- Review request messages for supermarkets
- Price and quote replies for supermarkets
- Delay and stock notices for supermarkets
- Thank-you messages for supermarkets
- Hand-over notes for supermarkets
- Ready answers for supermarkets
- Holiday and closure notices for supermarkets
- Missed enquiries calculator for supermarkets
- AI agent instructions builder for supermarkets
- Auto-reply message writer for supermarkets
- FAQ answer sheet for supermarkets
- Reply speed scorecard for supermarkets
- Is your business ready for an AI agent? Quiz for supermarkets
- Message workload planner for supermarkets
- Hours saved calculator for supermarkets
- Enquiry-to-customer value estimator for supermarkets
Read the guide: Answer shoppers' messages with an AI agent.