# EHR Extraction & Write-back ## Pre-Visit Chart Prep, Run as Agents Every chart on tomorrow’s schedule, prepped tonight: the lab result the visit depends on pulled forward, the hospital discharge summary found, the overdue screening flagged, before the clinician ever logs in. Chart prep is the most automatable job in the clinic, because it’s reading, not judging. Go-live in as little as 30 min · Browser agent · Customer portal login ## See it run on your systems. We map your process, volume, and exception paths, then recommend a practical first scope. ## Tomorrow's charts, prepped tonight instead of at 9pm Tomorrow's visit schedule identifies the charts that need preparation, Asteroid gathers each visit's results and outside records, runs your written prep protocol, and stages a per-visit summary with the gaps flagged by name, while anything needing clinical interpretation is flagged for the clinician, never resolved by the agent. ### Ambulatory EHRs ### How it actually runs 1. Sign in to the EHR with credentials from an agent profile and read tomorrow’s schedule. 2. For each visit, check the results it depends on: labs ordered since the last visit, imaging back or still pending, each named with its status. A visit scheduled to review a result that hasn’t arrived is flagged tonight, not discovered in the exam room. 3. Pull the outside records the chart is waiting on, discharge summaries, specialist notes, wherever the EHR receives them. 4. Run the chart against your care-gap checklist: screenings overdue, immunizations due, refills about to lapse, each flagged by name. 5. Stage a per-visit prep summary where your clinicians expect it. Where the protocol includes routine documentation entry, it’s entered exactly as provided and verified saved. The line is bright: a result that changes the visit, a contradiction in the chart, anything needing clinical interpretation is flagged for the clinician, never resolved by the agent. It moves information; it doesn’t practice medicine. 1. Open patient chart 2. Collect prior notes and results 3. Check approved prep rules 4. Identify missing items 5. Populate prep fields and tasks 6. A prepared chart, handed off before the visit Approved administrative fields populated, source material attached, and a visible missing-item list per visit, so the chart is ready because it's morning, not because someone gave up an evening. ## Pajama time is a staffing decision nobody actually made. No practice ever decided its physicians should do an hour of chart review at 9pm; the work landed there because no role owned it and only clinicians could tell what mattered. But the telling-what-matters part is your prep protocol, written down once. The finding-and-fetching part is reading, collecting, and arranging, the exact work agents absorb, and every hour of it a physician does is bought at physician cost and paid for again in burnout. The chart should be ready because it’s morning, not because someone gave up their evening. ## What runs today Chart-prep agents are built per EHR on request, and the machinery underneath them already runs in production: reading patient records out of eleven different EHRs at 40,000 extractions a month, and filing documents into charts with every upload verified. Prep is those two motions pointed at tomorrow’s schedule. ## Frequently asked questions ### Our prep checklist is our own. How does it become the agent's protocol? Your prep protocol is the configuration: which results get pulled forward for which visit types, which care gaps to check (screenings overdue, immunizations due, refills about to lapse), where outside records arrive in your EHR, and where the per-visit prep summary is staged so clinicians find it. Most practices have never written that checklist down, so scoping starts there, and each facility's EHR becomes a scoped build of the same protocol. Clinicians keep their morning routine; the difference is the chart is ready because it's morning, not because someone gave up an evening. ### What EHR access does chart prep require, and is it mostly reading? The agent signs in with credentials from an agent profile and reads tomorrow's schedule, then each visit's chart: labs, imaging status, discharge summaries and specialist notes wherever the EHR receives them. It's built per EHR on the machinery already running in production: record extraction across eleven EHRs and verified document filing. Prep is overwhelmingly read work; the only writes are staging the summary and, where your protocol includes it, routine documentation entered exactly as provided. Credentials stay scoped and revocable; security posture on the security page. ### What happens when a chart can't be prepped cleanly? The gap becomes the flag. A visit scheduled to review a result that hasn't arrived is flagged tonight, not discovered in the exam room; a pending lab is a named finding with its status, not a stop. Outside records the chart is still waiting on are reported as missing rather than papered over. Anything needing clinical interpretation (a result that changes the visit, a contradiction in the chart) is flagged for the clinician, never resolved by the agent. A chart the EHR won't open is a technical failure reported as unprepped, never presented as ready. ### How do clinicians verify and audit what the agent staged? Each visit gets a per-visit prep summary staged where your clinicians expect it, with every finding named: which labs are back or pending with status, which outside records were pulled, which care gaps flagged by name against your checklist. Where the protocol includes routine documentation entry, the entry is verified saved rather than assumed, and it contains exactly what was provided. Runs execute against tomorrow's schedule on a nightly cadence, so what was checked and when is inherent to the run record. A prep run whose main output is flags for the clinician is the workflow succeeding. ### Which prep and documentation steps can run automatically, and which clinical decisions stay with staff? The finding-and-fetching runs automatically: reading tomorrow's schedule, checking each visit's dependent results, pulling forward discharge summaries and specialist notes, and running your care-gap checklist. That work is reading, collecting, and arranging, not judging. The line is bright on the other side: deciding what a result means, resolving a contradiction in the chart, choosing what to do about an overdue screening, and anything else requiring clinical interpretation is flagged for the clinician, never resolved by the agent. Routine documentation entry runs only where your protocol explicitly includes it, entered exactly as provided. ### How do you prevent the agent from writing unsupported clinical content or into the wrong chart? Two rules, both structural. On content: the agent stages what exists and enters what it's given. Documentation is entered exactly as provided and verified saved, never generated, and anything ambiguous escalates rather than guesses, because documentation requiring judgment is a clinician's. On identity: prep is driven from tomorrow's schedule, each chart opened for its own visit, on the same chart-side machinery as verified document filing, where a wrong or ambiguous patient match stops the run. A wrong-chart write is the one error these workflows are built to refuse.