Agentic Index
Abridge vs Microsoft Dragon Copilot (2026)
Choose Abridge when every line of a note must trace back to the visit, and Microsoft Dragon Copilot, the successor to Nuance's DAX Copilot, when one platform should serve physicians, nurses and radiologists and host other vendors' agents. That verdict is the Agentic Index coverage score, graded from each vendor's own published materials.
Both listen to the visit, draft a note inside the EHR and hold it until a clinician reviews and signs. Abridge's Linked Evidence ties every sentence of the draft to the moment in the conversation it came from. It works in more than 28 languages across 55 plus specialties, and it runs inside Epic's Haiku, Canto, Hyperdrive and emergency module with notes syncing both ways. Its work reaches into structured orders, coding context for revenue cycle teams and prior authorization during the visit through Availity's payer APIs. Dragon Copilot drafts notes, orders, flowsheet entries, referral letters and visit summaries, with separate experiences for nurses on Epic Rover and radiologists in PowerScribe One. Since August 2026, US health systems can add coding, decision support, prior authorization and screening agents from a Dragon Copilot storefront in Microsoft Marketplace. Partners build on its REST APIs and SDKs, where Abridge offers no public API. Data location is the sleeper issue for systems outside the US. Abridge runs only in US data centers on Google Cloud, while Dragon Copilot keeps data in the customer's country across the US, Canada, the UK, France and Germany. Both are sold by quote. Abridge licenses each clinician by the year with no usage metering, and Dragon Copilot is quoted by Microsoft or a partner, with Marketplace agents billed separately on their own terms.
On the Agentic Index healthcare agent ranking, neither Abridge nor Microsoft Dragon Copilot clears the bar, which asks for all five chart loop capabilities documented in full. Abridge does not document knowledge grounding and RAG in full; Microsoft Dragon Copilot does not document knowledge grounding and RAG in full, nor observability and auditability. 4 of the 69 vendors in the lane clear it. See the healthcare agent ranking
This comparison is published by Agentic Index, an independent agentic AI vendor research platform. Abridge and Microsoft Dragon Copilot are each graded against the same 14 capability Agentic Index taxonomy, from the vendor's own public materials under the Agentic Index verification standard, alongside 1,061 researched vendors. No vendor pays for placement and no vendor has reviewed this page. How this evidence is graded
Choose Abridge if
- Clinicians should trace any sentence in the note to the audio it came from before signing.
- Your health system runs Epic and wants the scribe in Haiku, Canto, Hyperdrive and the emergency department.
- Prior authorization should be worked against payer rules while the patient is still in the room.
- Visits happen in many languages, with more than 28 recognized.
- Leadership wants peer reviewed evidence, such as the Kaiser Permanente study of 1,306 clinicians in NEJM AI.
Choose Microsoft Dragon Copilot if
- Nurses and radiologists need documentation help alongside physicians.
- Patient data must stay in Canada, the UK, France or Germany.
- You want to add coding, decision support and screening agents from other vendors in one place.
- Your developers or partners need REST APIs and SDKs to build on the platform.
- Sign in should run through your existing Microsoft Entra ID tenant.
| Feature | A Abridge |
M Microsoft Dragon Copilot |
|---|---|---|
| Action & orchestration | ||
|
Integrations & Tool Calling Ability to connect agents to real systems through native integrations, OAuth-authenticated actions, custom tools, APIs, webhooks, or MCP-compatible tools. |
||
|
AbridgeIntegrations & Tool Calling Abridge is embedded across Epic Haiku, Canto, Hyperdrive and the ASAP emergency module as the first partner in Epic's Partners and Pals and Workshop programs, with bidirectional note sync and write back into the chart. It also has an athenahealth partnership, and Availity's FHIR native payer APIs are embedded for prior authorization. Sourceabridge.com/press-release/abridge-inside-emoryread 2026-09-29 |
||
|
Microsoft Dragon CopilotIntegrations & Tool Calling Dragon Copilot runs inside Epic (including Epic Rover for nurses), PowerScribe One for radiology and other partner EHR integrations, delivers generated summaries directly into the EHR, and admins manage connectors in the Dragon admin center. Sourcelearn.microsoft.com/en-us/industry/healthcare/dragon-copilot/aboutread 2026-09-29 |
||
|
Workflow Orchestration Ability to sequence, branch, retry, route, and combine deterministic workflow nodes with autonomous agent steps. |
||
|
AbridgeWorkflow Orchestration Fixed function pipelines chain capture, note generation, structured orders, coding context for revenue cycle, and prior authorization with payer determination during the visit. Abridge sets that sequence, and customers do not configure it or combine several agents into a workflow of their own. Sourceabridge.com/press-release/abridge-availity-collaboration-announcementread 2026-09-29 |
||
|
Microsoft Dragon CopilotWorkflow Orchestration Beyond note drafting, organizations deploy AI apps and agents (coding, decision support, risk adjustment, prior authorization, screening) that run inside the Dragon Copilot point-of-care workflow alongside its own generation, and admins manage them per environment, so multiple agents work in one clinical workflow. Sourcelearn.microsoft.com/en-us/industry/healthcare/dragon-copilot/extensionsread 2026-09-29 |
||
|
Triggers & Channel Coverage How agents wake up and where they work: schedules, webhooks, message events, CRM events, inbox events, chat, email, voice, and collaboration tools. |
||
|
AbridgeTriggers & Channel Coverage Ambient capture is started by the clinician across iOS, Android and Epic desktop and mobile surfaces in outpatient, inpatient and emergency settings, including from the ED track board in Haiku, and speech models detect specialty, language and multiple speakers without manual adjustment. Capture happens inside the clinician's workflow, and nothing starts it on a schedule or from an outside event. Sourceabridge.com/press-release/abridge-inside-for-emergency-medicine-announcementread 2026-09-29 |
||
|
Microsoft Dragon CopilotTriggers & Channel Coverage Work starts when a clinician records an ambient conversation or dictates, from the EHR, the web, desktop or mobile apps, or asks Copilot a question. The clinician starts each of these, and there is no autonomous event, schedule or inbound trigger. Sourcelearn.microsoft.com/en-us/industry/healthcare/dragon-copilot/aboutread 2026-09-29 |
||
| Knowledge & context | ||
|
Knowledge Grounding & RAG Ability to ground agent behavior in company data through document ingestion, retrieval, external knowledge APIs, semantic search, or RAG layers. |
||
|
AbridgeKnowledge Grounding & RAG The Contextual Reasoning Engine integrates data from previous patient encounters, health system-specific guidelines and clinician preferences, context-aware evidence surfaces clinical insights from UpToDate, NEJM and JAMA contextualized to the encounter, and payer coverage requirements arrive through Availity's FHIR-native APIs for prior authorization. That context is gathered for each encounter and the literature is chosen by Abridge, and no store of the health system's own documents that the product searches is described. Sourceabridge.com/abridge-contextual-reasoning-engineread 2026-09-29 |
||
|
Microsoft Dragon CopilotKnowledge Grounding & RAG Copilot answers from the encounter transcript and notes, pulls EHR visit diagnoses into notes, and draws on trusted internal and third-party medical references. There is no retrieval structure the organization keeps over its own knowledge, and the reference sources are not named. Sourcemicrosoft.com/en-us/health-solutions/clinical-workflow/dragon-copilotread 2026-09-29 |
||
|
Memory & State Persistence Ability to persist context across a run, conversation, workflow, user, team, or longer-term memory layer. |
||
|
AbridgeMemory & State Persistence Editing tools adapt to clinician preferences, and the Contextual Reasoning Engine carries clinician preferences and data from previous patient encounters into each note. The personalization follows each clinician's style, and how long it lasts, or how a clinician reviews or deletes it, is not described. Sourceabridge.com/platform/cliniciansread 2026-09-29 |
||
|
Microsoft Dragon CopilotMemory & State Persistence Templates, custom prompts, shared and personal vocabulary and auto-style are settings and instructions the product applies, and finalized notes are the EHR's record. There is no agent memory with a scope and lifetime. Sourcelearn.microsoft.com/en-us/industry/healthcare/dragon-copilot/whats-new/5-0read 2026-09-29 |
||
| Control & trust | ||
|
Human Oversight & Guardrails Approval steps, consent checkpoints, escalation rules, structured guardrails, policy constraints, and pause/resume controls. |
||
|
AbridgeHuman Oversight & Guardrails Clinician review and sign off is a required step in the core workflow, since no one can commit a drafted note to the record without a clinician editing and signing it. Linked Evidence exists so the clinician can check each sentence against the source audio before signing. Sourceabridge.com/press-release/abridge-inside-emoryread 2026-09-29 |
||
|
Microsoft Dragon CopilotHuman Oversight & Guardrails All generated content must be reviewed and validated by the clinician before it becomes part of the medical record. Output is a draft for the clinician's review, with AI-generated sections marked and the source transcript one click away. Sourcelearn.microsoft.com/en-us/industry/healthcare/dragon-copilot/whitepapers/transparencyread 2026-09-29 |
||
|
Security, Identity & Governance RBAC, SSO, auditability, encryption, least-privilege tool access, compliance posture, and data handling policy. |
||
|
AbridgeSecurity, Identity & Governance The trust center lists SOC 2 Type 1 and Type 2 reports validated by an independent auditor for security and confidentiality, plus HIPAA, TX-RAMP and CCPA. Customer facing controls include audit logging, event and audit log management, single sign on, multi factor authentication, role based access, application and network penetration testing, and certificates of destruction, and a security whitepaper, network diagram, model cards and subprocessor list are published. Sourcetrust.abridge.comread 2026-08-30 |
||
|
Microsoft Dragon CopilotSecurity, Identity & Governance Users sign in through the customer's Entra ID tenant (OpenID Connect), and the Dragon admin center assigns user roles per experience plus Environment Administrator and Analytics Reader roles. Dragon Copilot is HITRUST CSF certified, with SOC 2 Type 2, ISO 27001, 27017 and 27018, FedRAMP, C5 and HDS listed. Sourcelearn.microsoft.com/en-us/industry/healthcare/dragon-copilot/whitepapers/securityread 2026-09-29 |
||
|
Observability & Auditability Traces, logs, execution histories, metrics, audit events, and debugging detail for production agent behavior. |
||
|
AbridgeObservability & Auditability Linked Evidence makes notes auditable by tying each AI-generated output to its source information across all input data, so every sentence of the note carries a record of what it was drawn from, and the trust center documents detailed audit logs of system activity, including access to and modification of sensitive data, kept in immutable storage. No run or step view of the scribe's work is described. Sourceabridge-contextual-reasoning-engine and trust.abridge.com; abridge.com/abridge-contextual-reasoning-engineread 2026-09-29 |
||
|
Microsoft Dragon CopilotObservability & Auditability Each encounter keeps its transcript beside the draft, and admin analytics report usage by site, session type and dropped users. There is no record for each encounter of the steps and actions the AI took. Microsoft's audit logs record access to its own environment, not the AI's actions. Sourcelearn.microsoft.com/en-us/industry/healthcare/dragon-copilot/whats-new/5-0read 2026-09-29 |
||
|
Deployment & Data Residency Deployment modes and options, including SaaS, dedicated cloud, VPC, on-prem, hybrid, local runtime, and self-hosting. |
||
|
AbridgeDeployment & Data Residency Abridge runs only as SaaS, hosted on Google Cloud with US based data centers, and offers no customer selectable region, no VPC, no on premises option and no deployment inside the customer's own tenant. It holds TX-RAMP authorization and keeps alternate processing sites for business continuity and disaster recovery. Sourcetrust.abridge.comread 2026-08-30 |
||
|
Microsoft Dragon CopilotDeployment & Data Residency Customer data is stored by the customer's location, within the EU Data Boundary for Europe, and never leaves the geography. US data stays in the US, Canadian data in Canada and UK data in the UK. Data from France, Ireland and Belgium is stored in France (EU), and data from Germany, Austria, the Netherlands and Switzerland in Germany (EU). Sourcelearn.microsoft.com/en-us/industry/healthcare/dragon-copilot/whitepapers/privacyread 2026-09-29 |
||
| Solution readiness | ||
|
Prebuilt Agents, Templates & Packs Ready-made workflows, packaged employees, templates, blueprints, industry solutions, and role-specific agents that reduce time-to-value. |
||
|
AbridgePrebuilt Agents, Templates & Packs Abridge ships note structures tuned per setting and specialty across 55 plus specialties and 28 plus languages, with purpose built variants for emergency medicine, inpatient and nursing documentation co developed with Mayo Clinic and Epic. These variants are built by Abridge, and there is no library of templates or agents for customers to install. Sourceabridge.com/press-release/abridge-inside-for-emergency-medicine-announcementread 2026-09-29 |
||
|
Microsoft Dragon CopilotPrebuilt Agents, Templates & Packs Organizations find, deploy and manage AI apps and agents from the Dragon Copilot storefront in Microsoft Marketplace, covering coding and documentation, decision support, risk adjustment, prior authorization, behavioral health screening and preventive care (US), beside separate physician, nurse and radiologist experiences. Sourcelearn.microsoft.com/en-us/industry/healthcare/dragon-copilot/whats-new/5-0read 2026-09-29 |
||
| Platform extensibility | ||
|
Model Flexibility & Routing Ability to work across multiple foundation models, route tasks to different models, or let buyers bring their own providers and keys. |
||
|
AbridgeModel Flexibility & Routing OpenAI and Google Cloud appear as data service vendors on the trust center's subprocessor list. No second model is named, no routing between models is described, and customers and admins cannot choose a model. Sourcetrust.abridge.comread 2026-08-30 |
||
|
Microsoft Dragon CopilotModel Flexibility & Routing The product uses LLM functionality and Microsoft's own speech and generation models, with Azure OpenAI classifiers named only for content filtering. There is no model provider list, and the customer has no choice of model. Sourcelearn.microsoft.com/en-us/industry/healthcare/dragon-copilot/whitepapers/transparencyread 2026-09-29 |
||
|
APIs, SDKs & MCP Extensibility Composability layer: stable APIs, SDKs, MCP tool consumption/serving, custom tools, and integration into internal systems. |
||
|
AbridgeAPIs, SDKs & MCP Extensibility No API, SDK, developer portal or MCP server for Abridge's own platform is published, and integration runs through native Epic embedding and partner connections. An API for licensed developers has been reported by a third party research profile, but Abridge does not describe one itself. Sourceabridge.com, platform/clinicians, abridge-contextual-reasoning-engine and trust.abridge.comread 2026-09-29 |
||
|
Microsoft Dragon CopilotAPIs, SDKs & MCP Extensibility The Dragon Copilot Developer Kit covers REST APIs for partners, SDKs for JavaScript, Android and iOS (DAXKit), an ambient audio streaming service and Entra ID access token requirements, and an extension framework (adaptive cards, configuration package, Marketplace offer) for building AI apps and agents on the platform. Sourcelearn.microsoft.com/en-us/industry/healthcare/dragon-copilot/sdk/get-startedread 2026-09-29 |
||
|
Testing, Debugging & Optimization Testing, debugging, scoring, retries, fallbacks, quality gates, and optimization loops for improving agent workflows before and after deployment. |
||
|
AbridgeTesting, Debugging & Optimization Abridge publishes model cards on its trust center, and an independent peer reviewed study with Kaiser Permanente across 1,306 clinicians, published in NEJM AI, evaluated the product. Customers get no testing, evaluation or regression tools of their own, and clinicians' corrections at sign off are the only feedback path. Sourcetrust.abridge.comread 2026-08-30 |
||
|
Microsoft Dragon CopilotTesting, Debugging & Optimization User feedback mechanisms feed continuous improvement, with nothing on how a change is measured for the customer. Microsoft's own pre-release accuracy tests, red teaming and A/B gating of new models against baselines test Microsoft's product. They give the customer no evaluation tools. Sourcelearn.microsoft.com/en-us/industry/healthcare/dragon-copilot/whitepapers/transparencyread 2026-09-29 |
||
| Specialist automation | ||
|
Browser & Computer Use Browser, desktop, or remote/local computer control for workflows that cannot be handled through stable APIs alone. |
||
|
AbridgeBrowser & Computer Use Operates through ambient audio capture and native EHR embedding with write back through Epic's own interfaces; no agent operates third party software through a browser or computer interface. Sourceabridge.com/press-release/abridge-inside-emoryread 2026-09-29 |
||
|
Microsoft Dragon CopilotBrowser & Computer Use The agent has no browser, desktop or computer control. Dragon Copilot captures audio and works through EHR integrations and its own apps. Sourcelearn.microsoft.com/en-us/industry/healthcare/dragon-copilot/aboutread 2026-09-29 |
||
Pricing snapshot
Sourced from the Index pricing dataset · open each vendor's profile for full detail.
| Pricing | A Abridge |
M Microsoft Dragon Copilot |
|---|---|---|
|
Entry price Lowest public entry point |
Contact sales. Abridge publishes no prices. | Not published. Dragon Copilot is sold through Microsoft and its partners. |
|
Pricing confidence How public the numbers are |
Contact only | Contact only |
|
Billing Primary billing axis |
Per clinician per year, as an enterprise seat license. | — |
|
Variable cost Workload / overage exposure |
Low variable cost | High variable cost |
|
Free tier / trial Try before you buy |
No free tier
|
No free tier
|
|
Buying motion Self-serve vs sales call |
Sales call | Sales call |
Dragon Copilot is a Microsoft product that succeeds the DAX Copilot and Dragon ambient tools from Nuance, which Microsoft owns, and it is sold through Microsoft and its partners without a public price. Abridge is independent, founded in Pittsburgh in 2018 and backed by Andreessen Horowitz and Khosla Ventures among others, with about $812 million raised. It negotiates contracts with each health system rather than listing prices.
More comparisons with Abridge or Microsoft Dragon Copilot
Other matchups in healthcare agents
Not the pairing you were after? These compare a different set of healthcare agents on the same 14 capabilities.