"DoctorAl: Build a unified LLM tool that would include optimization of routine tasks for clinicians, product and service recommendations, and handling CS tasks," the document said, likely referring to customer-service tasks.
The team said Amazon is in a good position to build this because it has first-party health data. That could be a reference to the rich data available through its One Medical primary care service and the Amazon Pharmacy prescription business.
The team also mentioned Amazon's "large clinical panel" as an advantage here. That may refer to a group of healthcare professionals that advises on relevant questions. This can also refer to the ability to perform a wide range of clinical tests.
The Amazon planning document also highlighted the company's existing homegrown AI models, such as the publicly launched Titan and the internally available Olympus. Big foundation models are often used as a starting point to develop more focused, specialized AI models and tools.
Amazon has been on an all-out blitz to stay relevant in the explosive generative AI space, after rivals like Microsoft, OpenAI, and Google raced ahead with early powerful models and AI services. Amazon CEO Andy Jassy has said that every team at Amazon is working on some kind of AI project, and the company is on pace to generate at least $1 billion dollars in revenue from those initiatives this year.
An Amazon spokesperson told BI that the administrative burden in healthcare is "immense," resulting in higher costs for customers and less time for providers to offer personalized care. One Medical already uses AI to automate certain workflows, and Amazon Pharmacy uses AI to deliver prescriptions to customers more quickly, the spokesperson added.
"Of course we're always looking for ways generative AI can be responsibly used to make the health care experience better for everyone," the spokesperson added.
This Amazon representative also stressed that the company does not use patient data to train AI models.
"Protecting customer privacy and security is foundational to how we design our AI products and services," the spokesperson said. "We do not use our customers' protected health information to train our current health AI models, and to imply we do would be incorrect."
BI previously reported on some of Amazon's internal AI projects, including a new AI-infused Alexa voice-assistant and a ChatGPT-like chatbot, codenamed "Metis." Amazon also created a new AGI team last year, led by Rohit Prasad, its head scientist and a senior vice president. Some of these businesses, like new AI chips built by Amazon Web Services, are struggling to grow in adoption.
For Amazon's healthcare team, it already had a number of other AI projects in the pipeline back in December 2023, when the planning document was created.
These initiatives included tools for optimizing administrative work through automated "support agents," likely in reference to AI bots that can handle complicated tasks on their own.
According to the document, some of the capabilities of these potential AI agents included:
In April, Amazon's chief medical officer Sunita Mishra shared some of the ways Amazon is thinking about AI in the healthcare space, including how to improve mundane administrative tasks.
Amazon expects these new AI tools to help reduce costs for its healthcare business. That's important for the company, because it's trying to cut hefty losses in this part of its empire, as BI previously reported.
The document said the team was looking for ways to "reinvent where and how AI and humans provide navigational and care services," without needing to increase full-service offices.
It also hoped to handle 87,000 customer service contacts in 2024 through "additional self-service content and the AI healthcare assistant service," the document stated.
"Customers will get help answering many health queries and navigating to the products, services, or solutions right for them via a free Al healthcare assistant service," the document added.
