Notable runs the clinical-AI surface inside the EHR. Dock holds the attributed review layer that sits between Notable's draft and the chart it touches. Each AI-generated note, order suggestion, or chart-closure recommendation lands in a Dock row, is read by a named clinician, and is signed off (or sent back) before anything propagates to Epic. The clinician's identity, the agent's identity, the model version, and the timestamp travel with the decision.
Notable and Epic stay the system of record for the raw data: the chart, the encounter, the order set. Dock is the system of record for what the AGENT INTERPRETS from that data. Each Dock row carries a pointer back to the platform record (notable_task_id, epic_encounter_id), the agent identity that produced the draft, the reviewing clinician, the decision, and the timestamp. The agent re-fetches Notable and Epic state via fresh API reads when it needs current chart context, never relying on a cached snapshot when the stakes are clinical. This is the pattern we describe across healthcare operations and the broader compliance posture work.
The Dock review surface
A single table, notable_ai_reviews, holds every AI suggestion awaiting a clinician.
| notable_task_id | epic_encounter_id | suggestion_type | agent | drafted_at | reviewer | decision | signed_at |
|---|---|---|---|---|---|---|---|
| NT-44918 | ENC-7720341 | A/P note draft | notable-scribe-v4 | 2026-05-30 08:14 | Dr. R. Okafor | approved with edits | 2026-05-30 08:21 |
| NT-44920 | ENC-7720388 | Order set: CHF discharge | notable-orders-v2 | 2026-05-30 08:17 | Dr. R. Okafor | rejected, manual entry | 2026-05-30 08:25 |
| NT-44923 | ENC-7720412 | HCC code suggestion | notable-coder-v3 | 2026-05-30 08:19 | Dr. M. Chen | approved | 2026-05-30 08:30 |
A worked workflow
Notable's scribe agent listens to an encounter and drafts an Assessment and Plan. Instead of writing directly to the Epic chart, the agent posts the draft to Dock, opens row NT-44918, attaches the encounter pointer, and records its own identity. Dr. Okafor opens the row, sees the diff against the prior note, edits two lines, and signs. Dock fires a consent-gated write back to Epic with Dr. Okafor's name on the signature line, not the agent's. The agent's contribution is preserved as a drafted-by attribution, separate from the signing clinician.
Order suggestions follow the same arc, but with a sharper gate. A CHF discharge order set is a dangerous operation: the clinician must explicitly approve, not merely acknowledge. If she rejects, Dock writes nothing to Epic, and the rejection becomes training signal for the next Notable model review.
Why it matters
Notable reports millions of automated tasks per day across its customers, with accuracy claims in the mid-90s for chart-review workflows (Notable). That is the right level of performance to draft, but not to sign. The AMA frames this distinction directly: augmented intelligence is meant to assist clinical judgment, not replace it, and physician oversight is the load-bearing principle for safe deployment (AMA).
Dock makes that oversight legible. Auditors do not have to reconstruct who reviewed what from log fragments. The review surface is a queryable table with agent identity on one side and a licensed clinician on the other. When the model rolls forward, the identity lifecycle record shows which version produced which suggestion, and the audit trail survives the upgrade.
The clinician keeps the pen. The agent keeps the draft work. The chart keeps a clean signature.
Try the Notable review template in Dock
FAQ
Does Dock replace Notable's review screens? No. Notable's in-EHR experience stays. Dock holds the cross-encounter, cross-clinician review queue and the attribution record. The two are complementary.
Does the patient chart change without a clinician? No. Dock will not write back to Epic unless a named, licensed reviewer signs the row. Auto-approval is not an available decision state.
What if the agent updates mid-shift? Each Dock row records the exact agent and model version that drafted the suggestion. Version rolls forward; old rows keep the version they were drafted under.
Is this compatible with our existing Epic audit workflow? Yes. Dock's write-back uses Epic's standard signature endpoints, so the Epic audit log sees a clinician sign, with a Dock reference ID for cross-lookup.
