Three years in: what we got right, what we got wrong

Three years ago today we shipped the first version of PhoneGPT to fourteen pilot plumbing shops in Texas. Some of what we believed then held up. A lot of it didn't.

Got right

• Trades is the right wedge. The pain is universal and the math is unambiguous.
• "Sounds human or it doesn't work." Every percentage point we gained on naturalness moved booking rates more than any feature.
• Text the owner, don't email them. Owners live in the SMS app.

Got wrong

• We thought owners would want a web dashboard. They wanted a daily digest text.
• We over-built the IVR. Nobody wanted menus. They wanted the bot to just talk.
• We under-priced for the first year. We thought we were selling minutes. We were selling jobs.

Thank you to the 12,400 shops who trust us with the most important thing in your business: the phone.