Your billing census runs on
one person's memory.
Over a day a week of retyping, at a five-facility network. The first real run composed 637 census entries with zero conflicts against the hand-typed ones.
First call is free. BAA before anything touches real data. A person approves every line.
What you end up with
Two screens. One for the person fixing it, one for the person funding it.
Shown with invented numbers so you can see the shape of it. Nothing below is a real client, and no patient data appears anywhere on this site.
Monday worklist
For the billing lead. Before claims go out.
Service after discharge
MR-5502 · Cedar Park · discharged the 2nd, services posted the 4th and 5th
Billed above documentation
MR-3871 · Northgate · IOP 3 hr on the census, no signed note in the chart
Under threshold
MR-4417 · Lakeview · IOP 2.0 hrs, below the 3 hr line
Level-of-care conflict
MR-2298 · Fairmont · stepped up to PHP, authorization covers IOP only
Unbilled attendance
MR-6130 · Northgate · attended and documented, never entered
Every flag is a candidate for review, not a verdict. A short session can be perfectly legitimate. The list exists so one person sees all of it at once, in the same week.
Executive summary
For ownership. No patient data, ever.
Census strength against break-even, month to date
Network census trend, with projection to month end
Cedar Park has run under break-even for three straight weeks. At this trend it closes the month nine percent short. All aggregate, so it travels to a phone without a single patient name.
Patterns you cannot see one week at a time
Repeat under-threshold attendance this month, by client reference
Any single short day looks like nothing. Six from one client in four weeks is not a billing problem, it is an engagement conversation for the site director. Nobody catches that reading one week at a time, which is exactly why nobody was catching it.
Not for fifty-site roll-ups, hospital systems, or anyone already live end to end on a modern RCM platform. If that is you, I will say so on the first call.
Why the numbers are wrong
Money leaves a manual census five ways.
Not fraud, not carelessness. Attendance lives in one system, money lives in another, and a person carries it across by hand every week.
The proof
Over a day of retyping became a review pass.
Five facilities, four outpatient. One operations lead. Roughly $500K a month in ad spend feeding admissions.
Anonymized · Behavioral health network
What changed
Before
- One person retyping the census for over a day, every week
- Seven people hunting discrepancies out loud every Monday
- Errors found after the billing company, or after a denial
- Four more facilities coming, scaling only by hiring
After
- The census fills itself. The ops lead validates instead of typing
- Each director arrives with their list. The meeting is a checklist
- Caught the same week, while the chart can still be corrected
- Every facility, same engine, same ten seconds. Headcount flat
The part that matters
You do not have to trust it.
Operators should not hand revenue to automation on faith, and I would not ask you to. The system proves itself against your own work before you rely on it.
It grades itself first
Everywhere your person already filled a cell, the engine composes its own answer and compares. That agreement rate is the trust number, and you see it before anything goes live. First real run at the network above: 637 entries, zero conflicts.
Then you run them side by side
Your person does one normal week by hand. The system does the same week. You read every agreement and every difference. The diff decides, not me. At that network, the client proposed the parallel run themselves.
How it works
From first call to live in under two months.
The audit reads your actual data and tells you where the money goes. The build turns those findings into the engine and the two dashboards. Nothing is guessed, because the audit is what scopes the build.
The Ops Audit
Two to three weeks
Your real census, not a demo. Ends with the leaks quantified and a billing SOP your team can run without the person who normally does it.
The Engine
Four to six weeks
Your rules go into config: rates, tiers, thresholds, and each site's spreadsheet dialect. It fills the census in your exact format and builds both dashboards.
Live
Then it just runs
Scheduled overnight, your list waiting Monday morning. Rules recalibrated as payers move, new facilities onto the same engine.
Why it goes this fast
Nothing gets replaced. No new platform, no migration. It reads the spreadsheets and logs you already keep, in the formats your sites already use, and writes into the census your billing company already reads.
The slow part of most software projects is change management, and there is almost none here. The people doing the work keep doing it in the same place. What changes is that they review instead of retype.
Pricing is flat and quoted on that call. If the audit finds nothing much, I tell you that, you keep the SOP, and we build nothing. That is a good outcome.
Questions
What operators ask first.
We already have a billing company.
So did they. Your billing company bills whatever you send them. This makes sure what you send is right. Theirs liked it, because we made their input cleaner.
Our data is too sensitive for AI.
Agreed, which is why patient data never leaves your infrastructure. Local scripts, BAA signed, aggregates only. Ask any vendor where the data physically goes. My answer is nowhere.
We are too messy, and every site does it differently.
Messy is the qualification, not the disqualification. If your spreadsheets were clean you would not need me. Each site has its own dialect of colours, time slots and naming, and the parser learns all of them before anyone talks about standardising.
How do we know it works?
Parallel run. Your person does one normal week by hand, the system does the same week, and we compare line by line before you rely on it.
What if our billing person leaves?
That is the real risk today, and a reason to start rather than wait. The rules come out of their head into config, plus an SOP we write together. The system runs the rules and a human confirms the edge cases.
Who does the work?
I do. Sunny Binjola. The person writing the code is the person on your calls.
Start with the call.
Thirty minutes, free, nothing to prepare. Walk me through how a session becomes a paid claim and I will tell you where I would look first. If you are not a fit, I will say so on that call.