Case studies · South Florida
Three operations,three weeks of work returned.
Each build started the same way: a walk through the operation, a list of the work still done by hand, and one thing fixed properly before anything else was quoted.
Representative engagements. Client names are withheld and figures are typical of a build of this size. Named references are available on request.
01/Marine refit and service
Quote to invoice, without the Friday reconciliation
A 40-person refit yard was re-keying every approved quote into accounting by hand. Now it posts itself and the yard closes the week on time.
- Weekly admin removed
- ~14hWeekly admin removed
- Job cost visibility
- Same dayJob cost visibility
- Manual ledger entries
- 0Manual ledger entries
- Where
- Fort Lauderdale, Broward County
- Timeline
- Mapped in week 1, live in week 4, handed over in week 6
- Built
- Custom job record · Accounting integration · Mobile change orders · Cost reporting
The situation
Estimators wrote quotes in one system, the yard worked from printed job sheets, and accounting re-typed the approved lines into the ledger at the end of the week.
Nobody could answer what a job had actually consumed until that reconciliation was done, so change orders were argued from memory and margin was a monthly surprise.
Two people spent most of Friday on it. Neither of them was hired to do that.
What we built
- 01
One record per job
Quote, work order, parts draw and invoice all read from the same record, so the yard and the office are never describing different versions of the same job.
- 02
Approved quote posts itself
An approval writes the job into accounting with the right cost codes, and flags anything it cannot match instead of guessing.
- 03
Change orders in the field
Techs raise a change order from a phone at the boat, with a photo, and it lands against the job before the customer has left the dock.
“The part that changed the business was not the automation. It was that everyone finally argued from the same number.”
02/Property management
Maintenance intake that answers at 2am
1,900 units across 30 buildings, one inbox, and a triage queue that only moved during office hours. An AI intake agent now opens, classifies and routes every request as it arrives.
- To first response, any hour
- < 2 minTo first response, any hour
- Fewer vendor call-outs
- ~35%Fewer vendor call-outs
- Queue instead of six inboxes
- 1Queue instead of six inboxes
- Where
- Miami-Dade County
- Timeline
- Mapped in week 1, live in week 5, second building group added in week 9
- Built
- AI intake agent · Work order system · Vendor routing · Resident notifications
The situation
Residents reported problems by phone, email and three different portals. Everything landed in one shared inbox that a coordinator worked through each morning.
Emergencies waited in the same queue as a squeaky door, so the genuinely urgent items were found by luck rather than by process.
Vendors were dispatched by whoever answered, which meant the same trade was called three times for one building in a week.
What we built
- 01
One intake, every channel
Phone, email, SMS and the portals all write into a single queue with the unit, the building and the resident already attached.
- 02
AI triage with a hard rule underneath
The agent classifies and drafts the work order, but anything reading as water, gas, electrical or access goes straight to the on-call number without waiting for a human to agree.
- 03
Vendor dispatch by building
Open jobs group by building and trade before dispatch, so one visit clears four tickets instead of four visits clearing one each.
“We stopped hiring for the inbox. The coordinator now works the exceptions, which is the job we actually wanted filled.”
03/Multi-site specialty clinics
Insurance verified before the patient arrives
Six clinics were verifying benefits by phone on the morning of the appointment. Verification now runs the night before, and the front desk starts the day knowing which visits have a problem.
- Returned to the front desk daily
- ~2hReturned to the front desk daily
- Instead of morning of
- Night beforeInstead of morning of
- Sites on one process
- 6 of 6Sites on one process
- Where
- Broward and Palm Beach
- Timeline
- Mapped in week 1, first clinic live in week 4, all six by week 10
- Built
- Payer integration · Scheduling sync · Exception queue · Audit trail
The situation
Front desk staff called payers each morning to confirm eligibility for that day's schedule, which took the first two hours and was never finished before the first patient arrived.
Anything unverified became a billing problem three weeks later, when it was somebody else's job and much harder to fix.
Each clinic had built its own workaround, so no two sites could cover for each other.
What we built
- 01
Overnight eligibility run
Every appointment for the next day is checked automatically against the payer, and the results are on the schedule before anyone opens the building.
- 02
An exceptions list, not a report
The front desk sees only the visits that need a call, with the reason and the script attached, instead of a spreadsheet of everything.
- 03
One process across six sites
The same flow runs at every clinic, so staff can cover another site without learning a different set of workarounds.
“The denials we used to find in week three, we now find the night before. That is the whole difference.”
Want to know which of these your operation is sitting on?
Free AI Opportunity AssessmentThe offer
Free AIOpportunity Assessment.
We walk your operation, find the work still being done by hand, and tell you which of it is worth automating first. You keep the list whether or not you build with us.
- Takes
- 30 minutes
- Costs
- No cost, no obligation
- You get
- A written opportunity list
- Back in
- Three business days
01
Where the hours actually go
A short walk through the steps somebody repeats every week, and what each one costs you in time.
02
What AI can hold, and what it cannot
An honest split. Some of this work should be automated, some should be fixed by changing the process, and some should be left alone.
03
A build order with rough numbers
The opportunities ranked by return, with an estimate of hours saved and what each build takes to stand up.