Thoth Labs · Confidential
BMET Copilot
The AI documentation copilot for biomedical equipment technicians.
A spoken sentence becomes a compliant service note — with the right
part number — in under a minute.
● 100% on-premise capable
Working product, in use today
Confidential
How it works
From a spoken sentence to a closed work order.
🎙️The tech just says: "Replaced the door latch on the warmer, self-test passed, returned to service."
🎙️1
Tech speaks
Plain language, the way they already describe the job.
🧠2
AI structures it
Problem, action, testing, ESD, outcome.
🔩3
Note + verified P/N
Matched to live inventory: stock, bin, cost.
📄4
Compliant PDF
Joint Commission–formatted record.
📤5
Upload to RSQ
Filed in RSQ, ready for the tech to close.
⚡ Under a minute, hands-free — note and part lookup run in parallel
The tech reviews, edits any field inline, and the finished note uploads straight to RSQ. No retyping, no separate part search, no new workflow to learn.
Live Demo
A real, working product — not a concept. We'll take a spoken sentence and
produce the finished, RSQ-ready note in front of you, end to end.
Ready when you are.
▶ Run app.thoth-labs.org
Where it fits
Two layers. One workflow.
The BMET workflow doesn't change. Two AI layers slot into it at the moments they add the most — and they don't overlap.
WO arrives
→
Diagnose
→
Repair
→
Document
↑AIQ surface
knowledge retrieval
↑BMET Copilot surface
documentation generation
Complementary. Not competitive.
Built for hospital IT
Your data never has to leave.
🔒 100% on-premise
Runs on a TriMedx-controlled machine inside your own network; no device or vendor data leaves your infrastructure; built to pass your security review.
🔑 Inherits your identity
Single sign-on and MFA through your existing identity provider — the same Okta system TriMedx already uses. No new logins to manage.
⚙️ Flexible, already-approved engine
Runs on the Claude TriMedx has already rolled out enterprise-wide, or on a dedicated VPS to save on token costs. The LLM layer is one you've already approved.
📋 Audit-ready, RSQ-ready
Every record is standardized for Joint Commission and uploads straight into RSQ.
Patient data is never involved — at all. BMET Copilot documents devices, parts, and procedures, not people. That's the line generic AI tools can't claim: the full productivity gain, with nothing sensitive leaving the building.
Why BMET Copilot
One copilot. Wins for the tech, and for the company.
⏱️
More capacity, no new headcount
~8 minutes of typing becomes under a minute; that time goes back into wrenching.
📄
Stronger documentation
Complete, consistent, audit-ready notes on every work order, not rushed one-liners.
📊
Better capital planning
Clean failure history shows when to repair vs replace, where the real money is.
BMET
Copilot
💰
Saves money across the operation
Accurate parts ordered once, fewer vendor calls, cleaner audits; it compounds, it's not one line item.
🎓
Faster onboarding
New techs inherit clear records and step-by-step guidance, not cryptic notes.
🤝
Complementary with AIQ
Slots beside the AIQ agent: documentation, not overlap.
📣 And it's what people want. In TriMedx's own poll, the top AI priority was less paperwork for providers — the #1 answer by far. This is the project employees are already asking for.
"Why can't we just build this with Claude?"
Claude is the engine. We built the vehicle.
Anyone can rent the engine. BMET Copilot is the domain data, the workflow, the field mapping, the quality controls, and the privacy architecture that make AI actually usable for biomed documentation.
| Dimension | BMET Copilot | Claude / general AI alone |
|---|
| Domain data | Built in — 148-part inventory, assets, work orders, PM checklists | None — tech must paste or know everything each time |
| Output format | Standardized note — 11 required sections mapped to CMMS fields | Free-form prose; format varies every time |
| Ease for the BMET | Fill fields / check boxes — zero prompt skill needed | Requires prompt-writing skill to get good output |
| Quality control | Field-level confidence flags + flags MISSING fields | No structured flagging; tech must self-check |
| Parts lookup | Two-layer filter against real inventory | Manual; no part numbers unless tech supplies them |
| Electrical safety test | Integrated ESD / EST capture | Tech must remember to include it |
| Record output | One-click PDF | Copy-paste, reformat manually |
| Data privacy (PHI) | Local LLM option (on-prem) — data stays in-house | Cloud by default — PHI leaves the environment |
| Consistency | Same quality regardless of which tech uses it | Depends entirely on the individual's prompting |
| CMMS integration | RSQ two-way integration on roadmap (Phase 6) | None |
"Claude is the engine. We built the car — by an actual BMET, refined over months. You can't prompt your way to that."
Let's talk
We're curious how you see it.
Your honest reaction — and your questions.
Thoth Labs · BMET Copilot