Should You Let AI Answer Your Customers?
A one-hour way to test AI on customer emails before any customer sees a reply.
Episode 3 is live. Listen below, read the episode summary, and use the free companion playbook.
Where to listen
About this episode
Andrew and June gave Gemini and ChatGPT a fictional cleaning-business fact sheet and eight made-up customer emails, once with no rule and once with a strict rule. The strict rule made both tools say they would check an unknown pet-safety question. On the broken-vase complaint, neither referred the claim to the owner with the strict rule; without the rule, both did. Two tools, one run each, made-up emails, one person grading. This is a small comparison, not a measured reliability rate.
Key takeaways
- Use AI to draft customer replies, not to send them on its own.
- Anything about money, safety or blame goes to a person.
- A strict rule made both tools say they would check an unknown pet-safety question.
- On the broken-vase complaint, neither tool referred the claim to the owner with the strict rule; without the rule, both did.
- Two tools, one run each, made-up emails, one person grading: this shows what to watch for, not how often AI gets it right.
Companion playbook
The step-by-step guide that goes with this episode. Free, no email required.
Should You Let AI Answer Your Customers?
A one-hour way to test AI on customer emails before any customer sees a reply.
Updated October 4, 2026
Read the playbookLet AI help me do this: one prompt, a fictional practice run and the checks
Show notes & links
- Should You Let AI Answer Your Customers? — the step-by-step companion guide for this episode.
Transcript
Put it to work
Optional: GreatArrow
Andrew also builds GreatArrow, a memory layer for AI tools. It's optional. Everything in this episode works without it.
GreatArrow.aiSubmit a problem
Got a problem in your business you'd like us to tackle? Send us one sentence at andrew@manito.ai.
Companion playbook: Should You Let AI Answer Your Customers?