Fifteen processes already run themselves. Here is what they cost us to build, what they are worth, and the four parts of this company they do not touch yet.
Every figure from the PSA Job List export of 2026-08-25 · 2,199 jobs · and from the Control Tower run at 07:05 this morning.
The organization
This is how a restoration company divides its work. Every job we take touches most of them, in roughly this order.
The Control Tower has been organized this way since the day it was built, on 14 August — before this meeting existed.
The organization
15 automated processes, across 6 departments. Nobody starts them. Nobody watches them. They report for themselves.
Four departments have no automation at all. That is not an oversight in the drawing — it is the plan.
What already exists
From the first file to this morning. Nine projects, built alongside a full-time job.
The Control Tower came last, and deliberately. You do not build the thing that watches everything until there is something worth watching.
What already exists
Yesterday, today, and tomorrow, without anybody logging on.
What already exists
10 of 15
rows the Control Tower grades as weak confidence — it can prove a step ran, not that its output was delivered. It volunteers that about itself.
A monitoring page that only ever showed green would be worth nothing. This one is built so that missing evidence is never reported as health.
That is the reason to trust every number that follows.
It takes us
135
days to get paid
The industry norm for a restoration contractor is 45 to 60.
Every extra day holds about $8,633 of our own money.
The accountants call it DSO — days sales outstanding. It is what we are owed, divided by what we bill in a day: $1,166,503 ÷ $8,633 = 135. It appears by that name later, so it is worth naming once here.
The problem
Across 101 jobs. $390,722 of it is over ninety days old — past the point where a carrier receivable stops being a payment and starts being an argument.
And it concentrates: Definity alone is $413,094 across 25 jobs — 35% of everything we are owed.
The problem
Collections are not broadly broken. A tail of files nobody owns is dragging the whole average — and that changes the fix. If everything were slow we would need more collectors. Because the tail is slow, we need something that never forgets a file.
The problem
Of our 548 open jobs, this is how many actually carry each date in PSA.
The record collapses exactly where the money is. Our recorded time-to-first-contact is 90 hours. Nobody believes that, and they are right not to — the date simply does not get entered. The honest finding is worse than a slow number: we cannot prove our response times to a carrier who asks.
The problem
First answer
$440,316
“completed work never invoiced” — every job finished with no invoice date on it.
After checking one job
$113,416
The largest supposedly-unbilled job in the company was already paid in full last November. Only the date field was blank. 69 of the 79 had already settled.
I am showing you the wrong number on purpose. It is the same disease as the slide before — 88 jobs that have received money carry no invoice date at all — and it is why every figure in this deck traces back to a file you can open. $113,416 across 10 jobs is real, and we can collect it this month.
The plan
Not in order of what is interesting to build.
Each stage ends with a new row in the Control Tower — so the system itself, not me, reports whether the stage landed.
The plan
DSO 135 days. $1,166,503 outstanding, $390,722 of it over 90 days.
The arithmetic, so you can check it: 60 days × $8,633 a day = $517,980. Not new revenue — our own money, arriving sooner. The collections cockpit already exists; it lost its data feed on 25 June and has been dark since. This is a reconnection, not a build.
The plan
65 claims where an adjuster asked us something and nobody replied. The oldest has waited 60 days.
Zero adjusters waiting over 3 days
lands mid October · Documentation
240 jobs inspected with no estimate written. Median 39 days since inspection.
Inspected → estimate: 380h → 72h
lands early November · Estimating
$331,315 of cost overrun, on the 39% of jobs where cost is recorded at all.
Realized margin 38.8% → 45%
lands early December · Intake
Stage 2 drafts adjuster replies but never sends them. On the first real run, twelve of thirteen auto-proposals would have answered a genuine question wrongly. The system drafts; a person sends. That is the design, not a limitation.
The plan
These dates are built on a rate we have already measured — nine projects and 15 unattended processes in 99 days, solo, alongside a full-time job. They are not a hope. They assume that same throughput, and they slip if a bad week at work takes one.
The same people, pointed at the work that actually needs a person.
Every hour on the left is spent moving a number out of one system and into another. None of it is judgement. The judgement — what to tell an adjuster, whether to escalate a carrier, whether a scope is right — stays with people, and they get more time for it.
Not asked for: headcount, new software licences, or a system replacement. Every stage builds on PSA, QuickBooks, XactAnalysis and Encircle — tools we already pay for.