Daily report · Tuesday 18 August 2026
The day began with Ken’s Sunday evening meeting and a list of fifteen jobs. It ended at midnight with four test calls to the 888 and a phone that got the price right, the package right and the inspector right. In between: the sewer scope mystery solved, a second meeting, two new voices, sixty-six tester cards, Ken’s own board, and a decision paper on our CRM. Nine sets of changes went live.
Josh used to offer a sewer scope and radon. GC doesn’t sell either one. He even made up a price. We had asked him not to, in writing, three times. It kept happening. So today we stopped asking and changed the code instead. Now he simply can’t say it.
He also has a new voice, leads with the 5 Star package, and says Greg will do your inspection — because Greg will.
One thing we learned the hard way: Chad’s half-finished setup form is what Josh reads from. It beats anything we write in the code. Three of today’s problems came from that one fact.
Dil brought in Sunday’s meeting transcript and asked what was done and what was not.
We went through the transcript line by line and matched it to what had actually shipped. Out of it came a clear list: what was done, what Ken had asked for, and the points where he had corrected us.
Two of his clarifications mattered most. “Stop marketing means stop marketing permanently” — so the Resume button is gone, and stopping now asks you to confirm because it cannot be undone. And “It’s Greg 9 and Greg at 1” — Chad no longer does the inspections.
Dil turned the meeting into a numbered list of fifteen. Most are now done. The ones finished today: the stop-marketing change, the scenario sheets, the schedule fix, six Josh fixes, the failover write-up, the fulfilment board, and the CRM paper.
Still open from that list: bot protection on the public forms. Not started, and not blocking anything.
Dil’s worry, in his words: “we cant launch this if brain 2 are leaking. do we have to answer the onboarding form for Chad and ted to stop the brain from leaking?”
Nothing was leaking. We checked: Quality doesn’t sell a sewer scope either, so there was nowhere for it to leak from. The $85 was GC’s own termite price, borrowed. Every fact Josh used was GC’s and correct — he had simply invented a service.
That mattered. A leak between two clients would have been a launch blocker and a confidentiality problem. This was one agent making something up, which is fixable.
Beth is buying the numbers for the people taking part. Out of that call came four new jobs, all done today: Josh sounds like Eric, graceful refusals, the green sample-data bar switched off on the demo, and compressed photos on two pages.
We also explained the two items Dil asked about — the Google Business Profile tick box and the AI disclosure — and both are now waiting on Ken’s ruling rather than on us.
Ken didn’t want a plain list. He showed us the version he liked and asked for that production value back — named characters, named houses, a play-it-this-way column, colours.
So we matched his design exactly, down to the teal, green and rust headings and the Consolas labels: 66 cards across three sheets, as Word documents, ready to hand to different testers.
A test run finished and the score looked bad. We read it call by call instead of trusting the number, and most of the failures were not real:
All three are fixed or written up. The one that stings: a scoring rule first reported on 28 July has been costing us points ever since.
Dil rang the test line four times. Nothing below is guessed — it all came off a real call.
GC doesn’t do sewer scopes. The $85 was Chad’s termite price, borrowed. Then Josh asked if the caller wanted to add it on.
We had a written rule against this. We had made it stronger that same morning. It happened anyway, twice in one hour.
So we stopped writing rules and wrote code. Josh now checks every sentence before he says it. If it names a service the inspector doesn’t sell, the sentence never leaves his mouth.
Josh said “We don’t do radon testing.” Correct. Then his very next sentence got blocked — because it also had the word radon in it.
The guard checks one sentence at a time. It couldn’t see that he’d just said no. So it kept cutting him off mid-answer, and the caller heard “let me just double-check that” five times in four minutes — twice in a row at one point.
Both fixed. The guard now remembers he already said no. And he can only say “one moment” once per answer, however many sentences get cut.
He said “the standard inspection is $450” and then offered the 5 Star at $572. Both numbers were correct. The order was wrong.
Dil’s point: the first number you hear is the one you compare everything to. Lead with $450 and the 5 Star sounds like an upsell. Lead with the 5 Star and $450 is the thing you drop to.
Odd part: Chad’s setup already said “offer the package first.” Josh was ignoring an instruction, not missing one. Now he leads with the package.
New voice. Led with the 5 Star. $572 and $450 both right for a 2,300 sq ft home. Turned down the sewer scope. Said the pre-drywall is $425 flat, not by square footage — and explained why. Said Greg Langston would do the inspection.
Asked “so where’s Chad?”, he said: “Chad Langston owns GC Home Inspection. He assists on inspections when needed, but Greg is the inspector who will be out for your pre-drywall.”
That’s a better answer than the one we wrote the fix for.
Chad started the online setup form and never finished it. It sits as a draft. We assumed a draft was ignored. It isn’t — whatever is in that form is what Josh tells real callers, and it overrules anything we write in the code.
So three separate fixes we made looked like they worked and reached nobody:
All three are fixed now, and we checked the live phone brain afterwards rather than trusting the change. But this will keep happening until Chad finishes the form. That is now the single most useful thing anyone can get from him.
Dil sent over Chad’s real ISN screens. They showed eight services for sale. Josh knew about three.
| Service | Price | This morning |
|---|---|---|
| Certified Move-In Ready | By square footage | Didn’t exist to him |
| Pre-listing / seller’s | Same thing, same price | He said no to it |
| Phase II inspection | $425 flat | Didn’t exist to him |
| Phase III inspection | $425 flat | Didn’t exist to him |
The one to get right: a Certified Move-In Ready is priced by size, like any normal inspection. It is not the $425 flat fee. Copying $425 across would have undercharged Chad on every big house.
He now sounds like Eric. Ken and Dil picked it. It was set in the code yesterday and didn’t take — a setting on the server was quietly overruling it. Removed. Parker on the website was moved to Liam at the same time, so the three agents are three different voices again.
On one call he said nothing for forty seconds. The caller asked “are you still there?”, then “is everything okay?”, then hung up.
There is a safety net for that. It never fired — because the caller’s own check-ins were resetting it. Every time he asked if we were alive, the rescue got pushed twelve seconds further away. Fixed.
Our automated tests had been calling Chad’s actual phone number — ten robot calls in one run. They now ring the shared test line instead, and the files are renamed so nobody picks the wrong one by mistake.
Ken asked for a fulfilment board — his tickets, separate from the engine work. It looks and works exactly like the pre-launch board. Same login.
It opened full of the pre-launch board’s items, which was our fault: a new board fills itself from a starter list, and we’d copied that behaviour across. Cleared, and it can’t happen again. Ken should refresh the page once.
Our automated call scoring has been marking correct behaviour as a failure. One of them was first reported on 28 July and has been costing us points on every run since. They can only be fixed in that system’s own screens, so they’re written up for whoever has the login.
Both showed up on the last call. Both are hearing problems, and both have a plan.
The caller said “it’s me, buying.” Josh heard a name in there and used it. There is already a written rule against this — it has failed four times now, so it needs code, like the sewer scope did.
Josh answered “we don’t do free inspections” — turning down a $425 job. He recovered a moment later, but a caller who took him at his word would have hung up. GC sells nothing free, and now sells pre-drywall and pre-listing, so this one mishearing can lose two different jobs.
Ken asked for the pros and cons of building our own CRM instead of using GoHighLevel, and asked for nothing to be built. The answer is: don’t build it — but not because there’s no time.
We have already moved six things off GoHighLevel since 5 August: the diary, the double-booking check, transcripts, recordings, the portal and client setup. Nobody ever called it a migration. The risk isn’t staying too long. It’s stopping something that’s working in order to do it all at once.
The full paper is linked below.
All checked live while this was written.
Worth knowing
The outage catch-up. Dil asked where the outage link is. There isn’t one — it isn’t a page. It’s a job that runs after an outage and picks up the calls we missed while we were down, so nothing is silently lost. Nobody opens it; it opens itself.
The tester cards. They are Word documents, not web pages — that was the point, so they can be printed and handed out. They came through this chat.
Nobody can do these for you