Fathom
Also known as: Fathom Health, fathomhealth.com
Autonomous medical coding platform that codes patient encounters directly from clinical documentation and returns coded charges to Epic, Cerner or athenahealth, charging only for encounters it successfully codes.
Fathom is an autonomous medical coding platform, founded in San Francisco in 2017 by Christopher Bockman and CEO Andrew Lockhart. It fuses deep learning with natural language processing to code patient encounters directly from clinical documentation, and unlike the broader revenue cycle platforms it competes alongside, it does one job across the widest possible specialty range rather than spanning intake, denials and collections.
Operationally it acts as a plug in. Patient information and clinical documentation arrive through EHR interfaces and APIs, the platform assigns codes, and coded charges return to the EHR for billing with minimal disruption to existing workflow. It supports Epic, Cerner and athenahealth, is listed in Epic's Toolbox programme under Fully Autonomous Coding, and is available through Google Cloud Marketplace. Beyond live production coding it handles backlog processing and provides real time audit.
The evidence base is unusually strong for this category and comes from third parties rather than the vendor. KLAS named it the number one solution for Reducing the Cost of Care in its 2025 Emerging Solutions Top 20 report, recorded a 95.5 out of 100 performance score, and validated automation rates above 90 percent with 100 percent customer satisfaction in a Spotlight report. A March 2026 deployment at Your Health across all service lines reported a 95.5 percent encounter level automation rate at 98.3 percent coding accuracy. Investors include 8VC, GV, Lightspeed, Founders Fund, Alkeon and Tarsadia, with a strategic investment from CVS Health Ventures in May 2026.
Two things a buyer should weigh. The commercial model is outcome aligned in a way that is rare here: pricing is per encounter with volume tiers varying by specialty, and Fathom charges only for encounters it successfully codes, so unhandled work costs nothing. Against that, independent research flags the material risk as disclosure rather than incident, with no lawsuits, breaches or regulatory actions surfacing but the underlying foundation model architecture left undisclosed.
Vendor details
Canonical URL
https://fathomhealth.com
Category
Healthcare agent
Company status
independent
Use cases & customers
Target customers
Deployment options
In practice
Your coding backlog is growing faster than you can hire certified coders. Fathom processes backlog alongside live production coding rather than requiring a separate remediation project.
You want AI coding without a department by department rollout. The platform covers a broad specialty range at once and plugs into Epic, Cerner or athenahealth rather than replacing the revenue cycle stack.
You are unwilling to pay for automation that does not land. Pricing is per encounter and Fathom charges only for encounters it successfully codes, so the unautomated remainder carries no fee.
Sources & related URLs
Research sources
Agentic Index coverage score
7.5 / 14 capabilities · 54%
| Integrations & Tool CallingPlugs into Epic, Cerner and athenahealth, receiving patient information and clinical documentation through interfaces and APIs and returning coded charges to the EHR for billing. Listed in Epic's Toolbox programme under Fully Autonomous Coding since August 2024 and available through Google Cloud Marketplace. 2026-08-05 | Full |
|---|---|
| Workflow OrchestrationRuns the coding task end to end without human handoff: ingests documentation, assigns codes and returns charges, across live production coding, backlog processing and real time audit as distinct operating modes. Validated automation rates above 90 percent mean most encounters complete unattended. 2026-08-05 | Full |
| Knowledge Grounding & RAGOutput is grounded in the encounter's own clinical documentation using deep learning and natural language processing, and coding rules are necessarily encoded, but no retrieval layer over coding guidelines, payer policy or a citable evidence base is documented. 2026-08-05 | Partial |
| Human Oversight & GuardrailsReal time audit gives coders a review surface and automation rates below 100 percent imply encounters route to human coders, but no confidence threshold, escalation policy or approval workflow is documented, and the commercial model implicitly determines what the platform declines to code. 2026-08-05 | Partial |
| Security, Identity & GovernanceHITRUST i1 Certified as of December 2024, satisfying hundreds of controls across 19 domains, alongside SOC 2 Type 2 compliance, strict HIPAA compliance and Business Associate Agreements, with security by design covering encryption, infrastructure security, real time threat management, access controls and workforce training. The company also argues its closed loop platform reduces exposure relative to outsourced coding, which involves many credentialed human users. 2026-08-05 | Full |
| Observability & AuditabilityReal time audit is a documented product capability giving visibility into coding output, and automation and accuracy rates are independently measured by KLAS, but no reporting dashboard, decision level audit trail or explanation of individual code assignments is documented. 2026-08-05 | Partial |
| Memory & State PersistenceEncounter history supports backlog processing and audit across time, but no agent memory, persistent client specific tuning state or cross session context model is documented. 2026-08-05 | Partial |
| Deployment & Data ResidencyCloud delivered as an EHR plug in, with Google Cloud Marketplace availability serving as a procurement channel rather than a deployment option. No self hosted, private cloud or regional data residency choice is documented. 2026-08-05 | Unable to verify |
| Prebuilt Agents, Templates & PacksShips the broadest specialty coverage in autonomous coding as prebuilt capability rather than per client configuration, across three packaged modes covering production coding, backlog processing and real time audit, and is listed as a ready to deploy option inside Epic's Toolbox. 2026-08-05 | Full |
| Triggers & Channel CoverageWork arrives as encounters through EHR interfaces and APIs, with batch backlog processing as a second mode, but no scheduled runs, event triggers or channel coverage beyond the EHR pipeline is documented. 2026-08-05 | Partial |
| Model Flexibility & RoutingNo model selection, routing or bring your own key capability, and independent research explicitly identifies the underlying foundation model architecture as UNDISCLOSED, naming it as a material buyer facing disclosure gap rather than an incident. 2026-08-05 | Unable to verify |
| APIs, SDKs & MCP ExtensibilityInterfaces and APIs carry documentation in and coded charges out, and Epic Toolbox and Google Cloud Marketplace listings extend distribution, but no public developer API, SDK or MCP surface is documented. 2026-08-05 | Partial |
| Testing, Debugging & OptimizationIndependent validation is exceptionally strong for this category: KLAS recorded a 95.5 of 100 performance score, validated automation above 90 percent with 100 percent customer satisfaction, and a March 2026 deployment reported 95.5 percent encounter level automation at 98.3 percent accuracy. But that is third party measurement rather than customer facing evaluation tooling, and no harness, benchmark suite or regression testing is offered to buyers. Graded partial to stay consistent with the dragon-copilot precedent. 2026-08-05 | Partial |
| Browser & Computer UseNo browser control or computer use capability. Work moves through EHR interfaces and APIs rather than by operating payer portals or other interfaces, which distinguishes it from adonis, where browser automations for claim status are documented. 2026-08-05 | Unable to verify |
The Agentic Index coverage score grades every vendor Full, Partial or Unable to verify against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Pricing
Not published; per encounter with volume tiers varying by specialty and encounter volume, billed monthly or annually
per successfully coded encounter, with volume tiers by specialty
What is public
The pricing structure and the charge only for successful coding term are publicly described; no rates are.
Billing mechanics
Per encounter pricing with volume tiers that vary by medical specialty and encounter volume, billed monthly or annually with a discount for annual commitment. The defining term is that Fathom charges only for encounters it successfully codes.
Cost watchouts
Encounters the platform does not code still need human coders, so the internal coding cost does not fall to zero and should be modelled alongside the per encounter fee
Variable cost rationale
Cost scales directly with coded encounter volume and is incurred only on successful automation, so spend tracks realised value rather than capacity, which is the lowest risk shape of consumption pricing in this index.
Additional watchouts
Independent research characterises the material buyer risk here as disclosure rather than incident: no lawsuits, breaches or regulatory actions surfaced, but the underlying foundation model architecture is undisclosed, which matters for a system making coding decisions that carry billing and compliance consequences.
Sales call required
Yes, required for paid access
Free / trial
No published free tier or self serve trial; evaluation runs through a sales process
Commercial notes
Charging only for successful automation is genuinely outcome aligned and rare across this index. It transfers the risk of low automation rates onto the vendor, which is a credible position to take only because independently validated automation runs above 90 percent. Buyers should still model the residual: whatever proportion is not automated continues to require human coders.
Key ambiguities
No per encounter rate, volume tier thresholds or specialty differentials are published, so cost cannot be modelled without a quote. The 30 to 50 percent savings claim, and the up to 70 percent figure, are vendor stated and not independently validated in the way the automation rates are.
Missing data
Per encounter rates, tier thresholds, specialty differentials, minimum volumes and contract length.
Related vendors
- Abridge — Enterprise ambient AI documentation platform, Best in KLAS 2025 and…
- Adonis — AI orchestration platform for revenue cycle management whose…
- AKASA — Generative AI for the healthcare revenue cycle with institution…
- Alaffia Health — Agentic AI platform for health plan claims operations that pairs…
- AlethianAI — AI native multi agent healthcare platform that automates clinical…
- Almanac Health — Research validated clinical AI platform whose Almanac Copilot acts…
Alternatives to Fathom
The closest documented capability profiles to Fathom among healthcare agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Sunoh.ai8.5 / 14Adds documented Deployment & Data Residency
- Alaffia Health9.0 / 14Fuller documented coverage on Knowledge Grounding & RAG and Human Oversight & Guardrails
- Arintra8.5 / 14Fuller documented coverage on Knowledge Grounding & RAG and Observability & Auditability
- FinThrive9.5 / 14Adds documented Browser & Computer Use
- Lyrebird Health8.5 / 14Adds documented Deployment & Data Residency
- Microsoft Dragon Copilot9.5 / 14Adds documented Deployment & Data Residency
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded