AI that works on patient data. Under a BAA.
Tally runs your books, dashboards, and month-end close on your practice's real data — the kind that touches PHI — inside a perimeter every vendor has signed a Business Associate Agreement for. The moment a task touches patient data, Tally will only send it to covered systems, and blocks everything else automatically.
BAA-backed, end to end
Every system in the path — AWS, the Claude models, Microsoft 365 — is covered by a signed BAA.
PHI stays in the perimeter
Once a conversation touches patient data, it can only send to covered systems — every other destination is blocked automatically, for the rest of that conversation.
U.S. AWS infrastructure
Compute, database, storage, and the Claude models (via AWS Bedrock) run inside U.S. AWS regions — encrypted, isolated per practice.
The difference
The work you can't safely do on ChatGPT or Claude.
Consumer AI tools aren't covered by a BAA, and they can send your prompt anywhere. Pasting a patient roster, an EOB, or a payroll file into one isn't compliant. Tally does the same work with the wall built in.
Consumer AI (ChatGPT, Claude, Gemini)
- • No Business Associate Agreement — PHI shouldn't go in.
- • Your prompt can leave to wherever the model provider runs.
- • No folder-level control over what data is in play.
- • No audit trail tying an action to an identity.
Tally by Ballast
- • BAA in place across AWS, the AI models, and Microsoft 365.
- • PHI-touched conversations can only reach covered systems.
- • Folders are labeled; reading patient data locks the session.
- • Every action logged to an identity for audit.
The Covered Zone
A walled garden for patient data.
Four controls run behind every action Tally takes. They're on by default, and once a conversation is carrying patient data, the block can't be bypassed — not by our team, not by yours.
Reading patient data locks the session
The moment Tally opens a folder that holds patient records, that conversation is flagged as carrying PHI — and stays that way.
It can only send inside the perimeter
Once carrying PHI, any attempt to email, search the web, or reach an un-approved app is refused unless the destination is BAA-covered.
No laundering into a “clean” folder
Tally won't write patient-derived content into a folder meant to stay free of it — closing the quiet-leak path.
A safety scanner as backstop
Even in a fresh chat, a Social-Security or card number in an outbound message is caught and stopped before it leaves.
What your practice can do
Real finance and operations work — on the data that matters.
Provider throughput reporting
Encounters per day, visits per FTE, new-patient mix, no-show rates — by provider and by site, every month.
Coding-quality review
E/M level distribution vs. benchmark, denial reasons, undercoding risk, and wRVU trends to protect revenue integrity.
Month-end close
Full close on your books with a reviewed package — while patient-linked detail stays inside the perimeter.
Payroll & benefits
Work with payroll and benefits data without it leaving a covered system.
Ask Tally anything
Chat with your books in plain English. Normal questions work as always; the wall only engages when patient data meets an outside destination.
Example output
The dashboards your practice gets.
Built from your own data and refreshed every month. Below are two examples — a provider-throughput view and a coding-quality view.
| Provider | Site | FTE | Enc/day | Vis/FTE | New pt | No-show | Status |
|---|---|---|---|---|---|---|---|
| Amara Okafor, MD | North Clinic | 1.0 | 26 | 372 | 14% | 6.1% | Strong |
| Ben Reyes, MD | Riverside | 1.0 | 25 | 355 | 12% | 6.8% | Strong |
| Marcus Bell, PA-C | Eastgate | 1.0 | 24 | 341 | 15% | 7.2% | On track |
| Chloe Tanaka, DO | Summit | 0.8 | 22 | 318 | 11% | 7.9% | On track |
| Priya Nair, CPNP | Lakewood | 0.6 | 20 | 296 | 13% | 8.4% | Ramping |
| Hannah Cole, CPNP | Westside | 1.0 | 15 | 214 | 18% | 9.1% | New |
| Provider | Encounters | Avg E/M | wRVU/enc | Accuracy | Denial | Undercode | Status |
|---|---|---|---|---|---|---|---|
| Amara Okafor, MD | 520 | 3.9 | 1.51 | 96.1% | 3.9% | 5.2% | Strong |
| Ben Reyes, MD | 498 | 3.8 | 1.44 | 95.0% | 4.4% | 6.8% | Strong |
| Marcus Bell, PA-C | 470 | 3.6 | 1.33 | 93.2% | 5.1% | 8.9% | On track |
| Chloe Tanaka, DO | 360 | 3.7 | 1.40 | 94.0% | 4.7% | 7.4% | On track |
| Priya Nair, CPNP | 296 | 3.5 | 1.28 | 91.4% | 6.2% | 10.6% | Watch |
| Hannah Cole, CPNP | 214 | 3.4 | 1.22 | 90.3% | 7.0% | 11.2% | Watch |
Illustrative example. “Bright Steps Pediatrics,” all providers, and all figures are invented for demonstration — no real practice or patient data is shown.
Bring your practice's books into the perimeter.
Ballast sets up and maintains the compliance layer for you. Tell us about your practice and we'll walk you through it.