Amazon Pharmacy Embeds Generative AI Models to Enhance Medication Distribution and Stocking
Amazon Pharmacy has begun incorporating generative AI into its digital medicine distribution service, employing large language models (LLMs) hosted by Amazon Web Services. The new AI features apply to almost every facet of the pharmaceutical business, including planning inventory and logistics internally, processing prescriptions submitted by patients, and engaging with customers asking questions or trying to find the best price for their medicine.
Amazon Pharmacy offers a platform for purchasing prescribed medication with the lure of the same speedy delivery of other Amazon purchases, including free, two-day delivery for Prime members. Amazon sees generative AI as a way to augment how that service functions in efficiency, accuracy, and availability. The platform relies on multiple Amazon Bedrock and Amazon SageMaker generative AI models that have been pre-trained for this purpose.
“Today, Amazon is using generative AI to make the pharmacy experience even better for our clinical teams and customers. We saw an opportunity to rethink information flows that was both exciting and transformational,” Amazon Pharmacy senior principal engineer Alexandre Alves explained in a blog post. “By building our solutions on Amazon Web Services, our teams were able to move faster and focus on creating the best experience for customers.”
The AI models can boost prescription order processing speeds by up to 90% while actually reducing human errors, according to Alves. The technology analyzes unstructured prescription data and transforms it into structured inputs to accelerate order fulfillment. Human pharmacists still double-check them, though, just in case. Internally, Amazon Pharmacy is leveraging generative AI models to forecast medication demand and also helps Amazon distribute the stock where it’s most needed.
For customers, Amazon Pharmacy is relying on generative AI-fueled analysis to offer real-time prescription pricing information, factoring in insurance coverage and ways to save through Prime and other programs. This makes comparison shopping on Amazon’s website feasible without requiring users to input insurance details manually. Amazon Pharmacy’s customer service will also get some generative AI enhancements. The models will review internal pharmacy documentation to generate summarized answers to common questions for staff. While the AI model provides the initial responses, Amazon’s clinical and customer service teams review all information for accuracy and safety before interacting with customers.
“Seeing the price of medications directly on a pharmacy’s website or mobile app is surprisingly novel,” Alves said. “LLMs can create immediate value for customers by surfacing up-to-date pricing, which enables better decision-making.”
This isn’t Amaozn’s first foray into healthcare powered by generative AI. AWS debuted a clinical assistant for healthcare providers last summer called AWS HealthScribe. The HIPAA-compliant service employs Bedrock’s LLMs to generate clinical documentation from doctor-patient discussions and reduce the time and energy medical care providers spend on paperwork and record-keeping.
Beyond Amazon, the Mayo Clinic has integrated Google’s Med-PaLM 2 LLM into its research hospital to help organize patient information, and Microsoft subsidiary Nuance started incorporating OpenAI’s GPT-4 LLM into the Dragon Ambient Intelligence platform medical professionals use to transcribe patient interactions back in March. And the British National Health Service is pushing hospitals to adopt generative AI tools with $27 million in grants. That’s before even considering pharmaceutical research, with human trials of a drug conceived and developed using generative AI underway.
“We are constantly discovering new uses for generative AI,” Alves said. “We’re excited about the potential to not only further improve the customer experience with Amazon Pharmacy, but also to significantly enhance the roles of our pharmacy staff, helping to create better health care outcomes.”