Automated payroll processing without the file surgery.
Someone hand-edits the scheduling-system export before every payroll run. We encode those rules, so the upload-ready file builds itself.
Official services partner of the platforms defining AI
NVIDIA
Anthropic
The export can’t be used as-is. Someone hand-edits it before every run.
From payroll and finance-ops leads we have sat with: healthcare staffing groups, facilities and field-services firms — dozens to hundreds of payees, not shopping for a new payroll system.
Hand-edited before every run
The payroll file from the scheduling system can’t be used as-is — someone hand-edits columns before every run; the goal is eliminating manual edits entirely.
Dozens of flags, fixed one by one
Every payroll run produces dozens of flags — missing facility setup, missing provider, missing payroll IDs — each one reviewed and fixed by hand.
The facility column lies
The export’s location column holds comma-separated facilities with the real one buried in parentheses, and typo variants break matching.
Rates live in someone’s head
Rates live in a separate system — or someone’s head; new providers can’t be auto-matched, so payroll IDs get pulled by hand each cycle.
Amendments arrive by email
Rate amendments arrive by email and ad-hoc meetings, then get applied manually — premium-pay deals for specific facilities live in inboxes.
Timesheets arrive by Dropbox
Timesheets and payroll reports are delivered by Dropbox and email, and the report layout changes when an outside vendor changes it.
When the payroll is contractor payroll.
What the manual way costs in medical payroll — healthcare staffing groups, payroll for home health care, any roster of contractors with facility-specific rates. Your numbers will differ; one cycle’s raw export puts figures on yours before anything gets built.
From anonymized engagements — healthcare staffing groups, dozens to hundreds of payees
We encode your exceptions.
No new payroll system to adopt. The edits your payroll person makes by hand become encoded rules; the payroll system stays the system of record.
- 01
Diff one pay cycle
We take one pay cycle’s raw scheduling-system CSV export and the hand-edited final file. The diff between them is the rulebook — yours to keep.
- 02
Encode the file rules
Facility-name normalization, rate tables per facility and call-vs-regular time, roster matching for new providers, amendments applied as they arrive.
- 03
Parallel-run until you trust it
The system reads each raw export and produces the upload-ready file, run in parallel against manual payroll line by line; exceptions queue for a human.
New payroll software won’t fix your file.
The flags come from the gap between your scheduling system and your payroll system — your facility names, your rates, your roster. The software holds the run; it doesn’t hold your rules.
We encode those; our AI does the reading against the raw export, your encoded rules do the judging — and the upload stops failing.
Run the new system in parallel with manual payroll, compare line by line.
Led by Peter Enestrom.
Asked by payroll managers and finance-ops leads.
The straight answers, before you book anything.
Automated payroll processing is the use of software to calculate gross-to-net pay, taxes, and deductions and to produce the payroll run without manual calculation. The term mostly refers to payroll software — the system that holds the run. In staffing-heavy operations the harder half is the input: turning the scheduling system’s raw export into a file the payroll system will accept. That preparation — reading the export, applying facility, rate, and roster rules — is what can be automated without changing payroll systems.
Yes. The payroll system stays the system of record; the automation sits in the gap between the scheduling system and the payroll run. Encoded rules — facility-name normalization, per-facility and shift-type rate tables, roster matching, amendments — are applied to each raw export, and the output is an upload-ready file plus an exception queue. No migration, no new payroll vendor. A scheduling system with no API is the normal case, not a blocker: a headless-browser scrape, a PDF parse, or an inbox read are all in-scope delivery patterns.
Because the file is assembled upstream of the payroll system, by hand. The flags — missing facility setup, missing provider, missing payroll IDs — come from the gap between the two systems: facility names that don’t match the setup, typo variants in the location column, providers not yet in the rate table, amendments that arrived by email after the last export. Nothing holds your rules, so each cycle’s payroll CSV regenerates the same mismatches, and someone clears them one by one before the upload will take.
Rates are encoded as a table, not remembered: per facility, and per shift type where it matters — call time versus regular time, premium-pay deals for specific facilities. Amendments are applied to the table as they arrive, including the ones that arrive by email. When a new provider can’t be matched against the roster, or a rate nobody encoded shows up, the line lands in an exception queue for a person instead of silently calculating wrong. This is the standard pattern in healthcare payroll services, where each payee may be a contractor with facility-specific rates.
Run it in parallel and compare line by line. The new system processes the same raw export alongside the manual payroll for one or more cycles, and every line is compared against the hand-built result until the outputs match. Exceptions are the point, not a failure: anything the rules can’t clear is queued with the rule behind it documented, so a person reviews the judgment calls. Trust is earned on matching output before the system touches a live run.
Start with one workflow.
Tell us where your team loses hours. We will come back with a straight answer on whether AI can help, what it would take, and what it would pay.