What was the result?
self-reportedThe Grout Guy founder Brad Young says an AI agent named Casey now closes about 40% of the company's email cases, which were all handled by hand a few months ago. At first its accuracy was about 70% and fixing its mistakes cost more time than it saved; the team added guardrails until it reached 95%, and anything it isn't sure about goes to a person.
The Grout Guy started as one man and a van and now has around 200 technicians across Australia. Every email comes into Salesforce as a case. An agent called Casey reads each one, and when it's a work order or quote request, a second agent called Walter enters it as a job ready to book. They built them with Salesforce's AI agent builder, where he says you now just prompt it, hook in the fields and tell it where to go. Brad still wants people talking to customers on the phone, and on the show he and the host noted that many callers hang up on after-hours AI phone bots.
Who did this?
How do I try it?
- Send all customer emails into one place, like your CRM, so each one becomes a case.
- Start an AI agent on one job only: sort emails and pull out work orders and quote requests.
- Measure its accuracy on real emails before letting it close anything.
- Add rules until it is right almost every time, and send anything uncertain to a person.
- Let it close only the simple cases, then widen its job step by step.
Editorial prompt, not The Grout Guy's setup. Read this customer email for [trade business]. Classify it as: quote request, work order, booking change, complaint, or other. Pull out name, address, phone and the job described. Rate your confidence from 1 to 10. If confidence is below 9, or it is a complaint, mark it for a person and do not reply.
Make it run every week
A play works once. Turn it into a routine in three small steps:
- Run it once and check the result yourself.
- Give your AI a schedule. Tell it when to run this play (every week, or every month) and to message you with what it found before you act.
- Keep a log. One line per run: what was checked, anything flagged, what you did.
What should I watch out for?
Self-reported. It took months of tuning to go from 70% to 95% accuracy, and Brad co-founded Servly, which sells this Salesforce setup to other trades. Safer version: keep complaints and anything unclear with a person, and spot check closed cases every week.
Did it work for you?
PageMotor keeps this site free.