How Amazon is approaching the healthcare AI race
The shopping and cloud giant is relying on One Medical and cloud dominance.
Morning Brew Design
The shopping and cloud giant is relying on One Medical and cloud dominance.
Morning Brew Design
As Big Tech races to dress its AI in scrubs, Amazon has a key advantage: It already owns hundreds of clinics staffed by human doctors.
Like some of its industry peers, the shopping and cloud giant has been rolling out a series of new AI-powered tools for providers and patients alike in recent weeks. Unlike competitors such as Microsoft and Google, however, these features are built around a real-world primary clinic network, One Medical, which Amazon bought in 2023.
“Ours is the only solution that is connected in the same app to an actual clinical delivery network,” Andrew Diamond, chief medical officer at Amazon One Medical, told us.
Amazon is counting on this acquisition, as well as its physical distribution network and its overall cloud dominance, to give it an edge when integrating the latest AI technology into healthcare. As part of an ongoing series on Big Tech’s race into the healthcare industry, we took a look at what sets Amazon’s strategy apart.
Earlier in March, Amazon Web Services (AWS) released a new agent platform designed specifically for healthcare administrative tasks, like scheduling, documentation, and patient verification. Around a week later, Amazon rolled out its virtual Health AI assistant for consumers to its website and app.
Over time, these patient-side and provider tools will become more interconnected, Diamond said. “We’re really in the early days of exploring [that],” he said. “Both of those sets of solutions exist inside the One Medical electronic health record, so they are connected at the level of the EHR that I use and my colleagues use every day, and we imagine that that connection will become deeper and more seamless over time.”
Diamond claimed that Health AI is already freeing up physicians to focus on patients with more complex issues.
“A very significant percentage of users of Health AI inside the One Medical app were getting their needs addressed without having to get to a clinician,” Diamond said. “A lot of people just have very simple questions, and they can get those questions answered.”
From One Medical, Amazon gets access to “a controlled clinical environment” and opportunities to experiment with AI without external providers, according to Mutaz Shegewi, senior research director of worldwide healthcare provider AI, platforms, and technologies at research firm IDC.
“One Medical is enabling Amazon to bring together a digital front door, telehealth, in-person care, and pharmacy as one to test AI across the patient journey,” Shegewi told us in an email.
But researchers also warn consumers to be wary about sharing private health information with AI chatbots, and studies of other AI chatbots dispensing health advice show a mixed record.
Diamond said Amazon uses an LLM-as-a-judge technique to evaluate every answer the chatbot gives and escalate flagged answers to human evaluators.
Prakash Bulusu, VP and CTO of Amazon Health, also said Amazon draws on its years of building out healthcare offerings when it comes to compliant data privacy.
“We have been in the Amazon Pharmacy business for a very long time,” Bulusu said. “We have gotten into the care business with One Medical and we had Amazon Care before that, so we have very high, set standards.”
Amazon’s other big edge is its general dominance in cloud services, which gives it a foothold in many health systems that are using AWS.
Whereas Microsoft has an eye toward workflow-level provider AI, through its Dragon Copilot ambient scribe, its enterprise software expertise, and a valuable longstanding partnership with Epic Systems, Amazon’s ambitions lie more in the underlying infrastructure, according to Shegewi.
“Amazon is looking more to power healthcare AI workloads at scale,” Shegewi said. “Amazon does operate and focus on the workflow level but is more focused on being the underlying AI substrate across the ecosystem. What sets it apart is its dominant cloud footprint in healthcare IT infrastructure, combined with mature AI platform services and logistics (and pharmacy) integration.”
But he also said that many healthcare providers would rather not be locked into a single cloud provider.
“Trust remains a gating factor for AI in general and especially as it pertains to hyperscaler expansion,” Shegewi said. “Large health systems are increasingly cautious about hyperscaler dependency. Multi-cloud strategies are common.”
That lock-in also means that the administrative healthcare AI features in the new Amazon Connect Health platform are only relevant to hospitals already in the AWS ecosystem, and cloud switching is difficult, according to Punit Singh Soni, founder and CEO of Suki, an AI healthcare startup that offers scribes and other administrative tools to providers. Soni said there’s also a question of whether the administrative healthcare features in Amazon Connect Health can outperform specialized startups like his.
“If you are a company whose only job is to do scheduling and patient engagement, and you’ve been doing it for the last 15 years, you have hooks into every aspect of the health system. You understand the data, you have all of that,” Soni told us. “Are you better placed? Or can Amazon, at a platform level, make this just a feature that somebody else can turn on?”
Ultimately, pairing its cloud services and other infrastructure at scale with the specific needs of healthcare workers will be key for Amazon, Shegewi said. “Amazon’s long-term success depends on how effectively it can bridge infrastructure scale with clinical workflow nuance.”
