Visit-close packets let clinicians approve referrals, orders, and follow-up work before the inbox becomes after-hours labor.

1. Heidi's $340 million financing, including a $240 million CVF commitment, shows that health systems now have budget and investor backing for supervised clinical-agent workflows rather than experimental note tools.
2. Multiple same-day sources say the category is moving from documentation into supervised action around patient visits, which makes visit-close workflow software newly urgent.
3. Order-set prepopulation, referral drafting, and pre-encounter record synthesis are already named as the next tasks under clinician review, so the first packet contents are concrete rather than hypothetical.
4. More than 10 million evidence queries and enterprise activation around 62 percent show that clinicians already use supervised agent assistance, making the next step turning answers into approved work.
5. Regulatory and clinical-safety foundations still gate rollout, and the UK and EU delay means U.S. systems need a clinician-in-the-loop pattern that can expand safely before broader international deployment.

Catalyst. Heidi's financing and enterprise traction show that ambient documentation budgets are already expanding into clinician-supervised action, with order sets, referrals, and pre-encounter records named as the next wedge.

Visit Close Packets sits after the note draft and before work fans out across the EHR. For each encounter, it compiles the relevant evidence, proposed orders, referral language, records requests, patient instructions, and staff tasks into one packet with risk tiers, required approvers, and downstream owners. The clinician can accept, edit, or reject the whole packet in one flow instead of opening multiple queues, while coordinators receive locked context and ready-to-complete tasks the moment sign-off happens. Every edit is captured by specialty, clinician, and action type so the product can shrink future reviews into lighter cards only when local acceptance-without-edit rates justify it. The result is less pajama-time admin for clinicians and fewer dropped handoffs for referral, scheduling, and records teams.

What's different. Ambient note vendors stop at documentation, referral-intake tools stop at one queue, and EHR inbox tools force clinicians to complete downstream work item by item. This concept owns the moment between note completion and operational execution by turning every next step into a single visit-close packet with required evidence and explicit downstream ownership. Its moat is the local acceptance telemetry: which specialties, task types, and evidence layouts consistently clear clinician sign-off without edits, allowing review time to collapse safely over time.

Jobs to be done

flowchart LR
  Visit[Completed visit] --> Packet[Visit-close packet]
  Packet --> Signoff[Clinician sign-off]
  Signoff --> Queues[Referral, order, and follow-up queues]
  Queues --> Outcome[Less after-hours work]

Executive takeaways

• The pain is current and budgeted: after-hours EHR work, documentation, alerts, and referral coordination still consume meaningful clinician and staff time, while health systems are already funding enterprise ambient AI to relieve that load.
• The mechanism is somewhat uncommon, not blank-space pure: public vendors expose note drafting, order capture, patient-message drafting, patient instructions, and EHR handoff, but none publicly lead with a vendor-neutral visit-close packet plus local acceptance telemetry across queues.
• Competition is real and getting closer. Abridge, Heidi, Epic, Microsoft-class copilots, Suki, and Nabla are all moving past transcription into broader workflow surfaces, so the wedge must be sharper than 'agentic clinical admin.'
• Reachable market is credible but narrower than a generic clinical-AI story: the first real beachhead is U.S. multi-hospital systems with employed ambulatory groups, ambient deployments, and centralized downstream ops.
• The concept is falsified if incumbents quickly bundle equivalent cross-queue sign-off into existing contracts or if pilots fail to reduce edit burden enough to remove, rather than add to, clinician review work.

Market definition

Service-enabled clinical workflow automation for U.S. health systems that converts ambient note output into signed post-visit execution across referrals, orders, patient instructions, and staff follow-up queues.

Customer and buyer

Primary operators are CMIO-led informatics teams, ambulatory operations leaders, referral or inbox managers, and service-line administrators in multi-hospital U.S. systems with employed clinics. Economic buyers are typically the CMIO, Chief Medical Officer, VP of Ambulatory Operations, or digital-health leader already accountable for ambient AI ROI and clinician experience.

Buying triggers

• After-hours documentation and follow-up burden remains visible even after ambient rollout, creating pressure to attack what still happens after the note is drafted. [1][3][4]
• Health systems are already making enterprise ambient-AI commitments, and expansion pressure naturally shifts toward orders, referrals, coding, and downstream workflow ROI. [5][6][7][19]
• Epic-native AI and similar tools are training clinicians to expect draft orders, patient instructions, follow-ups, and message responses inside the EHR, which raises the bar for post-visit execution software. [9][10]

Willingness to pay

Budget willingness is credible because health systems are already funding large ambient-AI deployments and expansion waves. Abridge reports use in more than 100 health systems, Heidi reports rapid ARR and enterprise activation, and Heidi publicly lists a $150-per-month clinician plan while teams and enterprise tiers remain custom sales motions. [5][6][7][19]

Category dynamics

Growth signal No clean public CAGR was retained, but vendor ARR, deployment, and usage signals imply category growth materially above standard health-IT norms.

Validation signals

• Abridge says it is deployed in more than 100 health systems and is expanding into broader clinical and financial workflows.
• Geisinger surpassed 1,000 Abridge users, reported about 45 fewer minutes in notes per 8 scheduled hours, and saw self-reported burnout fall in primary care.
• Heidi reports 175 million supported patient visits, about 62% enterprise activation, and an explicit move from documentation into supervised action.
• Epic publicly says Art can draft notes, orders, diagnoses, patient instructions, message replies, and referral context, confirming that post-note surfaces are already being productized.
• Referral coordinators in large primary-care operations already manage approvals, scheduling, EMR notes, and referral closure, proving the manual work surface is concrete and budgeted.

Regulatory & technical constraints

• ONC's HTI-1 rule requires predictive DSI transparency, ongoing maintenance, and risk-management practices for supplied predictive models.
• FDA CDS guidance keeps some clinician-facing support outside device scope but leaves device-policy obligations in place for software functions that remain regulated devices.
• The HIPAA Security Rule NPRM directly cites increased breaches, cyberattacks, and common compliance deficiencies, signaling higher buyer expectations around ePHI-handling workflows.
• Clinicians still must review, edit, and sign outputs, which limits how far orders, referrals, and patient instructions can move toward autonomy early on.
• Destination queues, order entry rules, and referral-close workflows vary materially across organizations, increasing implementation work.

The market splits into four overlapping layers: ambient documentation platforms that want to move downstream, EHR-native AI suites that draft notes and adjacent actions inside one stack, integration-first vendors that expose structured data and EHR handoff primitives, and the existing manual coordinator stack that still closes referrals and follow-ups by hand. The proposed startup sits across those layers by claiming the visit-close packet as the unit of clinician review and operational release.

Why incumbents do not win by default

• Ambient AI platforms. They win note capture and are moving into adjacent actions, but their public positioning still centers on documentation, coding, or broad assistant narratives more than on a cross-queue visit-close release object with measurable acceptance telemetry.
• EHR-native AI suites. They have the deepest in-workflow placement, but the public feature set is still a collection of drafts, summaries, and single-surface automations inside one EHR rather than a vendor-neutral orchestration layer across downstream teams and queues.
• Manual coordinator and inbox staff stack. It remains the default substitute because it handles local exceptions, payer quirks, and clinic-specific routing, but it is labor-intensive, slow to scale, and keeps clinicians and staff buried in review and close the-loop work.

Visit Close Packets should start as a U.S.-only workflow layer for 5-15 hospital regional health systems that already deployed ambient AI but still force clinicians to finish referrals, orders, patient instructions, and follow-up tasks across separate queues after the note is drafted. The acute pain is that ambient documentation reduced note burden without eliminating after-hours visit-close work, leaving CMIOs and ambulatory operations leaders with a visible ROI gap and ongoing clinician burnout. The MVP should compile low-risk post-visit work into one clinician sign-off packet, route approved tasks to existing coordinator worklists, and record acceptance-without-edit by specialty and action type before expanding autonomy. This beachhead is narrower and faster to prove than a broad clinical-agent suite because the first proof must show lower clinician review time and fewer downstream clarification loops in one dominant EHR and two repeatable service lines, not generic AI breadth. Research supports an estimated "$515.0M" TAM, "$126.0M" beachhead SAM, and "$11.9M" initial SOM, but those figures rely on price anchors and employed-physician assumptions rather than disclosed category spend. The company can win if it becomes the vendor-neutral release object between ambient note vendors, the EHR, and referral operations before incumbents bundle equivalent cross-queue sign-off. The biggest disconfirming risks are that clinician edit rates stay too high for the packet to save time and that implementation remains so services-heavy that gross margins never normalize. Because the inputs do not disclose real customer budgets, the first 12 months must prove paid pilot demand, repeatable implementation templates, and acceptance-without-edit thresholds on low-risk packet elements.

Problem

• Clinicians still close referrals, orders, patient instructions, and staff follow-up across separate EHR queues after ambient notes are drafted, which keeps after-hours work and fragmented review alive.
• Referral and scheduling teams inherit partial drafts and missing context, so they reopen work, chase clarifications, and make ambient-AI ROI harder for ambulatory leaders to prove.

Solution

• Compile each encounter's draft referrals, order suggestions, patient instructions, records requests, and follow-up tasks into one evidence-backed packet with risk tiers and explicit downstream owners.
• Keep clinicians as final signers, route approved work into existing coordinator queues immediately, and learn which packet elements can move from full review to lighter sign-off as local acceptance-without-edit rates rise.

Why we win

• The wedge is narrower than a generic clinical agent pitch: one visit-close release object that sits between the drafted note and operational execution across multiple queues.
• Acceptance-without-edit telemetry by specialty, clinician, and action type can become a compounding moat that note generators and single-queue tools do not naturally own.
• Incumbents already draft pieces of the workflow, but public positioning still centers on notes, orders, or messages rather than vendor-neutral packet release plus downstream completion benchmarks.

Milestones

flowchart LR
  Wedge[Visit-close packet wedge] --> MVP[Low-risk packet MVP]
  MVP --> Proof[Lower review time and fewer clarification loops]
  Proof --> Expansion[More service lines and benchmark moat]

Founding team

Experiment roadmap

Risk assessment

What must be true

• At least 3 of the first 10 qualified health systems will fund a paid pilot instead of waiting for Epic or their ambient vendor roadmap.
• Low-risk packet elements in primary care and cardiology can reach at least 60% acceptance without edit within 60 days.
• Median clinician review time per eligible visit can fall below 90 seconds while clinicians remain the final signers.
• Approved packets reduce downstream clarification loops or referral close delays enough to create budgeted ROI for ambulatory operations.
• After the first 2 deployments, implementation can be templatized enough to preserve target gross margin above 70% at scale.

Open diligence questions

• Which packet elements—referrals, order suggestions, patient instructions, records requests, or staff tasks—clear review fastest in one primary care and one cardiology clinic?
• How often do target buyers prefer an ambient-vendor expansion versus a third-party packet layer when both options are available?
• What baseline metric moves the budget owner fastest: after-hours EHR time, referral close rate, clarification loops, or coordinator productivity?
• Which EHR and ambient combinations dominate the beachhead, and how reusable are implementation templates across them?
• What governance artifacts do CMIO, compliance, and legal committees require before approving packetized order and referral drafts?

Model sanity

• Revenue engine. Base-case revenue comes from 7 paying health-system logos by Q4Y3, with the first 3 design partners converting into 6 production systems and 4 of those expanding to higher-value multi-service-line contracts.
• Must go right. The first two deployments must push acceptance-without-edit above 60% and cut implementation time materially so gross margin can climb from pilot-like 35% to above 70%.
• Model breaks if. The model is most fragile to sales-cycle delay, because a roughly one-quarter slip cuts about $255K of Y3 revenue and about $369K of ending cash in the sensitivity table.
• Next-round proof. The seed story is 2 production conversions, 60% no-edit acceptance on low-risk elements, 50% lower review time, and visible implementation-hour compression on the first EHR and ambient template.

Scenarios

Sensitivity

flowchart LR
  Targets[Qualified target systems] --> Assessments[Workflow assessments]
  Assessments --> Pilots[Paid design-partner pilots]
  Pilots --> Production[Production packet subscriptions]
  Production --> Expansion[More clinics and service lines]
  Expansion --> Revenue[Revenue]
  Revenue --> GrossProfit[Gross profit]
  GrossProfit --> Cash[Cash runway]

Flags: The plan only reaches 7 paying systems by Q4Y3, so one slipped health-system procurement or go-live meaningfully changes annual results. · Gross margin clears the long-run 70% target only if implementation time per added service line falls materially after the first two deployments; if services stay bespoke, the downside case compresses cash to about $0.2M. · Production ACV assumes partial service-line coverage rather than full-system rollout, so expansion inside each landed system is still required before the research SOM is relevant. · The business is still only near EBITDA breakeven in Y3, so the next round depends on proof milestones and not on profitability alone. · Cash is modeled as EBITDA and excludes payment timing, deferred revenue, capex, and any follow-on financing.

• Workflow variability. Each health system configures orders, referrals, and downstream staff queues differently, which can slow onboarding and reduce repeatability. Mitigation: Start with two or three repeatable service-line templates on common EHR and ambient-vendor patterns, then expand only after packet schemas stabilize.
• Clinician trust drag. If packets still require too many edits, clinicians will experience the product as another inbox rather than as real time savings. Mitigation: Launch with low-risk task types and use acceptance-without-edit thresholds before shrinking reviews or expanding autonomy.
• Vendor encroachment. Ambient note vendors or EHR vendors could add their own visit-close task bundles once the workflow proves valuable. Mitigation: Stay vendor-neutral across note sources and own the acceptance and downstream-completion benchmark data that single-vendor tools cannot assemble easily.