Product walkthrough
How tomorrow's schedule moves through the agent.
Six steps — from the schedule export to a scrub-tech-ready plan and supply-chain hand-off, all generated overnight without a human opening an app. Every decision the agent makes, with the reason behind it.
The agent ingests tomorrow's full schedule overnight.
The agent pulls tomorrow's full OR schedule from your scheduling system's nightly export. No human triggers the run. No live system access required. The agent reads the case list, the surgeon assignments, and the patient clinical characteristics — and begins generating plans without anyone opening an app.
Tomorrow · 7:30 AM
OR 12 · Estimated duration 4h 15m
Dr. Hollis
CABG ×3
Elective · On-pump · LIMA to LAD, SVG ×2
62y · BMI 27 · ASA 2 · EF 55
No prior cardiac surgery · Elective
The agent generates a plan for every case autonomously.
For each case, the agent's deterministic engine weighs the de-identified case against a hard safety layer that overrides anything touching patient safety. Every recommendation carries the reason behind it. The agent does this for the entire schedule before anyone arrives.
FIG. 01 / SCHEMATIC
Engine input/output flow
PARE OS v1.0 / RULES-BASED PHASE
Reviewed: 2026-05-25 / [Attending TBD]
The agent's output: a tier and a reason for every item.
Open Now items are opened before the case starts. Hold Sealed items stay in the room, unopened, available if needed. Leave Off items don't come to the OR at all. Each item carries the clinical or behavioral reason for its placement, so scrub techs and surgeons can interrogate any recommendation in seconds.
The scrub tech walks in to a reviewed, ready plan.
At 6 AM, the scrub tech sees the agent's plan for their cases — financially stripped, no costs or waste projections, just the clinical recommendation. They can mark items used, hold items unused, or request additional items mid-case. Every action is logged and feeds back to the engine. The agent is the prep, not the authority — the scrub tech and surgeon are.
PULL LIST · SAMPLE
Dr. Hollis · CABG ×3 · OR 12 · 7:30 AM
Every completed case sharpens the agent.
When the case closes, the outcome — what was used, what was wasted, what was requested mid-case — updates the surgeon's behavioral profile. The agent recalculates probabilities for that surgeon's next case. Over weeks, the recommendations get tighter. Over months, the per-case waste drops. No manual reconfiguration. No vendor required to update preference cards.
FIG. 02 / SCHEMATIC
Continuous learning loop
PARE OS v1.0 / CONTINUOUS LEARNING
Cycle: 24h / Batch: nightly
Overnight, the agent flags shortages to supply chain.
After generating every plan, the agent compares tomorrow's demand against on-hand inventory and par levels. Items below par get flagged to the materials manager with a suggested substitution or reorder — before the first scrub tech arrives. The supply-chain hand-off is part of the same autonomous overnight run, not a separate workflow.
FIG. 03 / HAND-OFF
Supply-chain notification
PARE OS v1.0 / OVERNIGHT RUN
Trigger: post plan-generation
Request a shadow analysis.
Send six months of de-identified cardiothoracic case data — procedure, surgeon, patient clinical characteristics, items opened per case. We return a CFO-grade waste analysis in fourteen days. No commitment, no BAA required.
Safe Harbor de-identified data is not Protected Health Information under HIPAA (45 CFR §164.514(b)(2)). Pilot deployment beyond the shadow-analysis stage operates under full BAA with the hospital and hosting providers.