Do and don't guide for your support AI agent
This guide gives plain do and don't rules for an AI agent answering support messages. It is for support teams who want the agent to sound human, stay honest and know its limits.
Runs in your browser. Nothing you enter is sent anywhere or saved.
Before you start
An AI agent follows the instructions you give it. Vague instructions give vague behaviour, while clear do and don't rules give you an agent that sounds like your team on a good day. Support is a place where tone and honesty matter more than speed alone.
Read each rule and keep the ones that fit your team. Reword any that do not sound like you, then paste the final list into your agent's instructions. Review the rules each month, because your products and policies will change over time.
Tick each one as you go. 0 of 10 done.
[ ] Do answer in short, warm sentences a customer can read on a phone [ ] Do say sorry once when something went wrong on your side [ ] Do ask for the order number early [ ] Do tell the customer what happens next and when [ ] Do hand over when the customer asks for a person [ ] Don't promise refunds, dates or exceptions a person has not confirmed [ ] Don't guess when you do not know the answer [ ] Don't argue with an upset customer or explain why they are wrong [ ] Don't give medical, legal or financial advice [ ] Don't message anyone who has not written to you first
Make the rules your own
Rules work best with an example. Beside each one, add a good and a bad sample reply from your own chats. The agent, and any new teammate, will understand the idea far better than from a bare rule.
Add a few of your own rules about words you like or dislike. If your brand never says unfortunately, put that in. Small touches make the agent sound like you.
- Keep the list under fifteen rules so it stays easy to follow.
- Test the rules with three tricky practice chats.
- Update them whenever a chat shows a new gap.
FAQ
How do I give these rules to my AI agent?
Paste them into the instructions your agent uses, in plain words, one rule to a line. Your agent follows written instructions, so the clearer and shorter they are, the better it behaves.
What if the agent breaks a rule?
Read the chat, find which rule was unclear and reword it with an example. If it keeps breaking a serious rule, narrow what the agent handles until you are confident again.
Should the same rules apply to my human team?
Mostly yes. Honesty, warmth and not promising what you cannot deliver suit everyone. Your team can bend some rules with judgment, which is exactly why handovers matter.
More free tools for customer support teams
- First-reply checklist for customer support teams
- Hand-over rules checklist for customer support teams
- First-week checklist for customer support teams
- Weekly chat review checklist for customer support teams
- Busy-season checklist for customer support teams
- Customer privacy checklist for customer support teams
- Greeting messages for customer support teams
- After-hours replies for customer support teams
- Booking and order confirmations for customer support teams
- Reminder messages for customer support teams
- Apology and complaint replies for customer support teams
- "Let me check" replies for customer support teams
- Follow-up messages for customer support teams
- Review request messages for customer support teams
- Price and quote replies for customer support teams
- Delay and stock notices for customer support teams
- Thank-you messages for customer support teams
- Hand-over notes for customer support teams
- Ready answers for customer support teams
- Holiday and closure notices for customer support teams
- Missed enquiries calculator for customer support teams
- AI agent instructions builder for customer support teams
- Auto-reply message writer for customer support teams
- FAQ answer sheet for customer support teams
- Reply speed scorecard for customer support teams
- Is your business ready for an AI agent? Quiz for customer support teams
- Message workload planner for customer support teams
- Hours saved calculator for customer support teams
- Enquiry-to-customer value estimator for customer support teams
Read the guide: Answer support messages with an AI agent.