Amazon Connect Health – Agent AI for healthcare

Amazon didn’t build a hospital. Build a stick. AWS recently introduced Connect Health, an AI platform for agents designed to handle the workload that burns out human workers more quickly. Editing. Clinical documentation. Medical coding. Patient verification. Medical history review. Five skills under one platform, and the first two are available now and the rest are being released in preview.

Time is not hidden. Health centers across the country are administrative staff that arrest. The burnout rate among medical coders and editors has been on the rise for years, and the workload on doctors continues to grow while the workforce does not. Amazon looked at that gap and decided that the fix wasn’t hiring more people. An AI was being created that could do a job without needing a lunch break or taking a burnout break.

What Amazon Connect Health’s agetic AI actually does

Each Connect Health capability targets a specific bottleneck in the healthcare workflow. The scheduling feature handles appointment booking with conversational AI, allowing patients to call and book directly without sitting on hold or leaving a voicemail. The clinical documentation feature listens to patient encounters and takes structured notes in real time, priced at $99 per month per provider for up to 600.

Medical coding capabilities translate clinical documents into billing codes, a process that is notoriously error-prone when done manually and expensive when done by professionals. The patient verification feature verifies a patient’s identity with HIPAA-compliant conversational AI before appointments, priced at $0.15 per verification action. The medical history review feature pulls patient records from all care settings and presents a brief summary before the visit, flagging active conditions, chronic problems, and trends that may be important to both treatment and billing.

What sets this apart from the AI ​​tools already floating around healthcare is the “agent” part. These are not chatbots sitting on the sidebar waiting for someone to ask a question. They are autonomous agents designed to perform multi-step workflows autonomously, only passing to humans when making a decision that requires clinical judgment. AWS positions them as co-workers, not tools.

The $99 question

The monthly fee for a clinical documentation agent is a number that will get a lot of attention, and rightly so. Clinical documentation is one of the greatest immersion periods in medicine. Doctors spend nearly two hours on paperwork for every hour of patient care, according to a 2016 study in the oft-cited Annals of Internal Medicine that has become the most pressing health care statistic.

At $99 a month, a literary agent costs less than one hour of a medical writer’s time in most markets. If we work as advertised, the numbers are so bad that acquisition becomes a question of trust, not budget. Can a healthcare provider trust an AI agent to accurately capture the nuances of a patient encounter? That’s the real barrier, and it’s one that no price point can solve on its own.

AWS describes a multi-step testing process for the document feature, which includes supervised fine-tuning, reinforcement learning, and clinical trials, but has not yet published public accuracy benchmarks. In healthcare, accuracy is not a nice thing to have. Missed information in clinical records can lead to coding errors, billing disputes, and in extreme cases, treatment decisions based on incomplete records. The cost is compelling. The story of accuracy still needs to be told.

Where Connect Health fits into Amazon healthcare

This isn’t Amazon’s first foray into healthcare, and it won’t be the last. The company shut down Amazon Care in 2022 after its primary care service, built for employees and later sold to employer health plans, failed to gain traction with business customers. It acquired One Medical for $3.9 billion in a deal that closed in early 2023, betting on primary care clinics. One Medical already uses Connect Health documents for more than a million visits, with an increase in medical codes this year. That makes One Medical a proving ground and a sales platform for what the platform can do at scale. Now Amazon is going after the infrastructure layer with Connect Health, and this approach feels very much in line with what Amazon does really well.

Amazon has always been much better at building platforms than running services. AWS is proof. Connect Health follows that playbook: instead of trying to be a healthcare provider, Amazon sells the tools healthcare providers use to work. It’s a game of picks and shovels for hospital administrators, and it leverages the one thing Amazon has that no other healthcare startup can match, which is the AWS cloud infrastructure that already runs within thousands of hospital systems.

Connect Health integrates with existing health record systems, removing the biggest friction of acquisition for business health consumers. It provides compliance with HIPAA and other healthcare regulations, which are table stakes but still worth checking given the sensitivity of the data these agents will handle.

Big picture

Amazon is betting that the health care management problem is a technical problem, not an operational one. That’s a simple framework for software for a retail company, but it’s not bad either. Paperwork has been overstaffing for twenty years, and hiring more workers isn’t going to fix it.

The setup of the Connect Health module is a smart move for an industry that hates change. Each skill covers one job, so the hospital can add scheduling without touching the documentation. A clinic can start with patient authentication and a coding layer later. You test one piece at a time, and you know exactly where to look when something breaks.

Whether it actually reduces burnout and saves money won’t be clear for months. What is clear now is that Amazon is no longer experimenting with healthcare AI. It sells it as a product line, with pricing and a release system designed to scale quickly.

Connect Health is live with AWS, although not all capabilities have reached general availability yet. Ambient Documents (GA) use a monthly payment of up to 600, and patient verification (GA) costs $0.15 per action. Appointment scheduling, patient information, and medical coding are currently in early testing, and prices for those capabilities have not been publicly disclosed. Healthcare providers interested in the platform can access it through their existing AWS accounts.