A denial arrives from a payer at 6:14 a.m. By 6:17 the agent has read the remit, pulled the chart note from Cerner, pulled the encounter from Epic, drafted an appeal letter, and parked it in a Dock row tagged for the revenue-cycle lead. The lead reviews at 8:00, edits one sentence, and approves. The appeal goes back to the payer with the lead's name on it and the agent's reasoning preserved. That is the shape of Dock for healthcare operations when Olive AI is the RCM substrate.
Olive AI, Cerner, and Epic stay the system of record for the raw data: the 835 remit, the encounter, the chart, the prior-auth thread. Dock is the system of record for what the agent INTERPRETS from that data. Every Dock row carries a pointer back to the platform record (olive_denial_id, cerner_encounter_id, epic_claim_id), the agent's identity, the drafted appeal, the reviewer's signature, and the timestamps on both sides. The agent re-fetches platform data via fresh API reads when it needs current claim status, because the platforms own current state and Dock owns interpreted state.
The Dock denial-appeal table
| denial_id | olive_event_id | epic_claim_id | denial_reason | agent_draft | reviewer | status |
|---|---|---|---|---|---|---|
| DEN-7821 | olv_evt_44a | clm_99812 | CO-50, medical necessity | Drafted MN appeal citing CDI note 2026-05-22, ICD-10 N18.6 | nora.s@health.org | approved 2026-05-30 08:14 |
| DEN-7822 | olv_evt_44b | clm_99813 | CO-197, no prior auth | Drafted retro-auth request with op-note excerpt | nora.s@health.org | needs revision |
| DEN-7823 | olv_evt_44c | clm_99814 | CO-16, missing modifier | Drafted corrected claim with modifier 25, no appeal needed | auto-routed | resubmitted 2026-05-30 09:02 |
One workflow, end to end
Olive emits a denial event. The agent reads olive_denial_id, fetches the encounter from Cerner and the claim from Epic, classifies the denial code, and writes a draft appeal into a new Dock row. The row inherits the agent's identity and a pointer back to every source record. The revenue-cycle lead opens the row, sees the agent's cited evidence inline, edits the medical-necessity paragraph, and approves. Approval triggers a consent gate before the appeal is transmitted to the payer through Olive's outbound channel. The row stays in Dock as the durable artifact: who drafted, who approved, what the payer saw, when it shipped.
Why this matters
Initial claim denials hit 11.8% in 2024, up from 10.2% a few years prior, and hospitals lose an average of 4.8% of net revenue to denials, per HFMA's denials research HFMA, 2024. The administrative cost to rework a single commercial denial averages $63.76. Volume matters, but attribution matters more. When an appeal goes out drafted by an agent and approved by a named lead, the audit trail is intact, and CDI specialists can review root causes rather than rerun the same denial loop, which AHIMA identifies as the core of effective denials prevention AHIMA Journal, claims-denials archive.
Service accounts cannot carry this weight. A shared bot account drafting appeals is the same as no signature at all. Agent identity is the precondition for the row to mean anything to a payer, an auditor, or a CDI lead. Audit trails and compliance live in the Dock row, not in a log file no one reads.
Finance leadership cares about this because attributed denial work shortens the appeal cycle and shrinks the write-off pool, which is why this pattern shows up on the Dock for finance page as well as the Dock for compliance page. Same row, two audiences.
Set up your Dock + Olive AI workspace.
FAQ
Q: Does Dock replace Olive AI? No. Olive AI stays the RCM event substrate. Dock holds the agent's draft, the reviewer's approval, and the pointer back to the Olive event.
Q: What happens if the payer disputes the appeal? The Dock row is the artifact. It shows which agent drafted, which lead approved, and which Cerner and Epic records were cited. The revenue-cycle lead can hand the row to a payer auditor without reconstruction.
Q: Can the agent transmit appeals without a human? Only behind a consent gate, and only for low-risk denial categories (missing modifier, demographic correction) that the revenue-cycle lead has pre-approved. Medical-necessity appeals always route to a human signer. See agent identity for the inheritance model.
Q: How does this affect coder workload? Coders stop rerunning the same denial loop. The agent drafts, the lead approves, and CDI specialists get a Dock view of root-cause patterns instead of a queue of one-off appeals.
