Three Healthcare AI Agent Platforms Shipped in 10 Days

Amazon, Salesforce, and Microsoft each released healthcare-specific AI agent platforms between March 5 and March 12, 2026. Amazon Connect Health went generally available on March 5. Salesforce announced six new Agentforce Health agents the same day. Microsoft opened a waitlist for Copilot Health on March 12. Each targets a different layer of healthcare operations, and each locks customers into a different vendor ecosystem. The timing is notable given that healthcare AI agent adoption remains in single digits despite enormous administrative waste across the industry.

The speed matters less than what it reveals. Three of the largest cloud providers now have production-grade agent infrastructure for healthcare automation — AI patient scheduling, clinical documentation, claims processing, patient verification — and the federal regulatory framework designed to oversee these tools was built for traditional medical devices decades ago.

What Each Platform Does

Amazon Connect Health is the most operationally focused of the three. Built on Amazon's existing Connect contact center platform, it ships five AI agents covering patient verification, appointment scheduling, ambient clinical documentation, clinical note generation, and medical coding. At $99 per month per user, it is one of the first healthcare AI agent platforms with transparent, published pricing — a notable move in a market where most vendors negotiate enterprise contracts behind closed doors. The platform is HIPAA-eligible and integrates directly with electronic health record systems.

Salesforce took a broader approach. Its six Agentforce Health agents span referral routing, data exchange between EHR systems like Epic and Athenahealth, insurance claims handling, epidemiology analysis for public health agencies, an offline mobile app supporting telehealth automation for rural providers, and a hospital operations command center for staffing and equipment management. Amit Khanna, Salesforce's SVP and general manager for health and life sciences, framed the thesis plainly: "How can we give time back to clinicians and reduce time in paperwork?" The agents become available starting June 2026.

Microsoft's Copilot Health is the outlier. Rather than targeting provider operations, it focuses on the consumer data layer — using medical records AI to aggregate health information from wearable devices like Apple Watches and Oura rings, hospital systems, and lab results into a single patient-facing view. The platform connects to clinician directories across thousands of U.S. hospitals and routes health questions through physician-reviewed answer sets. Microsoft already handles a massive volume of daily health-related queries across its platforms. Copilot Health launched on a waitlist with no firm date for general availability.

The distinction matters for healthcare organizations evaluating these tools. Amazon is selling operational efficiency for provider workflows. Salesforce is selling coordination across health system functions. Microsoft is selling patient-facing data aggregation. None of them talk to each other natively, and an organization running Amazon Connect Health for scheduling alongside Salesforce for referrals would need to build its own integration layer — or find a platform-agnostic agent infrastructure that works across vendors.

What Early Deployments Show

Several health systems have reported measurable results from AI clinical workflows, and documentation is where the gains are most consistent.

Cooper University Health Care reported saving several minutes per patient visit on clinical documentation using Dragon Copilot — time that adds up across hundreds of daily encounters. Mercy saw nurses reclaim meaningful time each shift through ambient documentation, with measurable improvements in both documentation speed and patient satisfaction scores. Sentara Health recovered thousands of nursing hours through similar tools, redirecting that capacity toward direct patient care.

The more striking result comes from MUSC Health, where 40% of prior authorization requests are now completed autonomously by AI agents. Prior authorization has long been one of the most labor-intensive workflows in healthcare — dedicated staff navigating insurance requirements for every procedure, every referral. Automating nearly half of that volume represents a fundamental shift in how health systems allocate administrative labor. And one large health system using Amazon Connect Health reported freeing hundreds of hours of staff time per week from patient verification alone.

These results come from early adopters willing to invest and share. Deloitte's research suggests that a majority of healthcare executives are now either building or implementing agentic AI, with most planning to increase investment. But "building or implementing" covers everything from a single pilot department to organization-wide deployment. The production-scale results from Mercy, Cooper, and MUSC Health are still the exception.

The Regulatory Landscape

Healthcare AI regulation was not designed for autonomous agents that process prior authorizations, generate clinical notes, or route patient referrals. The current framework was built for a different category of technology, and it shows.

The FDA's 510(k) pathway — the primary federal route for clearing AI medical devices — approves products that are "substantially equivalent" to existing cleared devices. That pathway clears 98% of AI medical device submissions — a rate that sounds reassuring until you consider it means the approval process rarely pushes back. Penn Medicine faculty have called this into question. Eric Bressman at Penn Med has noted that the framework was "developed decades ago for traditional medical devices" and has "a lot of holes in it." His proposed alternative is a "supervisory model that leads towards graduated autonomy," mirroring the way medical training works — starting with tightly supervised tasks and expanding autonomy as competence is demonstrated.

Congress has yet to pass direct health AI legislation. The White House released a four-page AI framework on March 20 calling for federal preemption of state laws. The framework does not address healthcare specifically. States have filled the gap unevenly. Nearly every state introduced health AI legislation in 2025, and dozens of those bills became law — but the rules vary widely by jurisdiction. In 2026, Manatt Health is tracking hundreds more state AI bills, concentrated in four areas: mental health chatbots, patient disclosure and consent, preventing AI tools from impersonating clinical providers, and payer use of AI.

Harvard Law professor I. Glenn Cohen summarized the situation directly: "The vast majority of medical AI is never reviewed by a federal regulator — and probably no state regulator."

For healthcare organizations deploying agent platforms right now, compliance is a moving target. State rules vary, federal preemption could override them, and the federal rules that would replace state action have not been written yet.

What Healthcare Organizations Should Consider

The practical question for hospital CIOs and compliance officers is how to deploy platforms that are shipping today under a regulatory framework that is still forming.

Bressman's graduated autonomy model offers useful guidance. Start with verification and documentation — tasks where the agent assists rather than decides. Expand to scheduling and claims processing as confidence in accuracy and auditability grows. Defer clinical decision support until oversight mechanisms are more established.

The interoperability question is equally pressing. Each of these three platforms integrates with EHRs differently — Amazon through direct EHR hooks, Salesforce through MuleSoft and FHIR/TEFCA standards, Microsoft through its HealthEx data aggregation layer. An organization choosing one vendor's agent stack is making a platform commitment with real switching costs. The more workflows, training data, and operational processes get built on top of vendor-specific integrations, the harder it becomes to change course later. Healthcare providers already experienced this with EHR vendor lock-in over the past two decades — the same dynamic is now emerging with agent infrastructure.

AgentPMT's architecture addresses this lock-in directly. A model-agnostic agent infrastructure layer — with audit trails for every agent action, budget controls for automated spending, and cross-platform portability — lets healthcare organizations deploy agents from any vendor without betting their operational stack on a single provider. As agent autonomy grows beyond documentation and into prior authorizations and clinical coordination, comprehensive activity logging moves from a nice-to-have to an operational requirement.

The three platforms that shipped in March represent genuine capability for healthcare operations. The early adopter data shows measurable gains in documentation, verification, and claims processing. The regulatory gap means those deploying today are operating in gray space — legal enough, but without the clear guardrails that federal oversight would provide.

Healthcare organizations that start deploying now will need internal governance frameworks that go beyond what any single vendor provides. Clear audit trails for every agent action. Budget controls that prevent automated spending from scaling unchecked. And a vendor strategy that preserves flexibility — because the rules governing these tools will eventually arrive, and the organizations with documented agent behavior will be the ones prepared to meet them.

---

Sources

• AWS launches a new AI agent platform specifically for healthcare — TechCrunch
• Salesforce releases six new AI agents for healthcare — Healthcare Brew
• Microsoft launches AI platform, Copilot Health — Healthcare Brew
• What's the state of healthcare AI regulation? — Healthcare Brew
• White House AI Framework Pushes for Broad Preemption of State Laws — Governing
• Many health care leaders are leaning into agentic AI — Deloitte
• 'Holes' in federal AI healthcare regulation should be patched — Daily Pennsylvanian
• How Amazon is approaching the healthcare AI race — Healthcare Brew
• Amazon Introduces Agentic AI for Health Care Providers — AHA Market Scan
• What Frontier Healthcare Leaders Are Doing Differently with AI — Microsoft Industry Blog