Gen AI Case Study: Amazon Pharmacy | The AI Moment – Episode 13

Gen AI Case Study: Amazon Pharmacy | The AI Moment – Episode 13

On this episode of The AI Moment, we discuss a generative AI case study– Amazon Pharmacy.

Gen AI has overrun the planet in concept, but the fact is the practical use of generative AI is nascent. Amazon’s Pharmacy case study is an important one that reveals lessons enterprises should learn and a blueprint they should follow in their journeys to unlocking the value of generative AI.

Watch the video below, and be sure to subscribe to our YouTube channel, so you never miss an episode.

Listen to the audio here:

Or grab the audio on your favorite podcast platform below:

Disclosure: The Futurum Group is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this webcast.

Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of The Futurum Group as a whole.


Mark Beccue: Hello, everybody. I’m Mark Beccue, Research Director for AI with The Futurum Group. Welcome to The AI Moment, our weekly podcast that explores the latest developments in enterprise AI. We are literally in a moment. The pace of change and innovation in AI is unprecedented, and the world has never seen anything like what we’ve experienced since ChatGPT launched in late 2022 and kickstarted the generative AI era.

And so with this show, The AI Moment, we’re going to distill the mountain of information, separate the real from the hype, provide you with sure handed market analysis from the latest advancements in AI technology and the mutating vendor landscape to things like AI regulations, ethics and risk management. So each show is typically about 30 minutes long. Today, we have a case study and I’m going to be covering what I call gen AI case study for Amazon Pharmacy. So this is how this goes. I do a lot of reading, I’m out looking around at things all the time. And this past week on the 24th, Amazon posted a blog and it was a description of how Amazon is using generative AI to automate a bunch of processes and improve customer experience for Amazon Pharmacy. So I thought that was interesting to look at. And so what caught my attention was how this is such a great lesson to learn for most of us that are looking at enterprises that are looking to implement AI, particularly generative AI and kind of the pragmatic approach. The use cases interesting so I thought I’d share that with you today in our segment.

So here’s kind of the key details. Amazon Pharmacy is like a full service pharmacy within, the store. And they talk about what you can do there. Customers can use it to purchase medications prescribed by their doctors and have them delivered. And Prime members have this, some benefits like free two-day delivery and some savings. There’s something called an RxPass which offers access to unlimited eligible prescription medications for $5 a month. And all these Amazon Pharmacy customers have 24/7 access to a pharmacist to discuss questions about their medications. And that’s one thing.

In the blog post, the key person is Alexandre Alves, A-L-V-E-S. He’s a senior principal engineer at Amazon Pharmacy. And what is interesting, if you read through the blog post, so these are key points, like I said, he’s trained as a computer scientist and had pre generative AI experience working with AI models, deep learning models in the area of learning to streamline supply chains and all those kinds of things. So he’s the one that led the initiative. And the third big point here is too, they went through thinking about their generative AI capabilities. They went to AWS, which is where many tools and platforms live, and Alves and the team used what they described as multiple pre-trained models from two pieces from AWS, Amazon Bedrock and Amazon SageMaker. SageMaker’s the machine learning tool platform in AWS. And again, here’s some interesting things. The team is using gen AI right now to address several issues.

One was, the first one was how to streamline some manual elements to digital prescription filling. Second one was how to forecast medication demand and to streamline and speed up delivery. Third one was how to enable the clinical and customer care teams to answer medication questions faster. And the fourth one was how to deliver these medications to users more efficiently. Four different areas they worked on.

So let’s look at those four areas real quick. The digital prescriptions and what they were talking about, the problem in this case is that these digital prescriptions contain confusing, inconsistent language, particularly in the directions for the user. So workers have to sift through these prescriptions by hand to edit and clarify and confirm the data, which slows down the process of filling the prescriptions. So Amazon solved this challenge by running the original, it’s unstructured data through a generative AI model, which uses a process called Name Entity Recognition to create a structure for the text. And they use categories such as dose or frequency. And from there, that helped the Amazon Pharmacy clinical staff fill these prescriptions and provide more clear instructions for the patients. So the end result, when they ran these models and kind of looked at the unstructured data, the final prescriptions are still reviewed by a pharmacists who checked the AI to make sure that what they’re suggesting in terms of these directions work. Right?

So when they did this, so they combined this approach, this generative AI approach with the expertise of the pharmacists. And Amazon, this is an interesting point because it’s a data point for a metric for success. It was they said, Amazon has increased order processing speed by 90% and reduced the rate of human error. So it’s problem number one. They did this, they got that kind of result. The second one was around forecasting medication. And what that was is think about it as forecasting medication demand. And what they wanted to do was help them, Amazon, stock the right medications in each shipping location so it would be ready to be dispensed when a prescription came in. And what the team did in this case was they used generative AI to synthesize data to test different stocking scenarios, both for how much you would spend, how much you would send to a shipping location, but also for determining dispense methods on site. So they put them in these canisters, and they were looking for the most efficient way to do that.

Apparently that worked out pretty well as well. I don’t have a metric from that, but that’s what they used it for was to do that. And then third, they found that it was difficult in many instances to decide to fill an order individually or in batch, right? So that natural tension you have when, “Is it that immediate? Can we wait to fill this when we are doing things together or do we have to fill this right away?” So Alves said this, and this is a quote. He says, “By improving our ability to ingest data and interpret context, generative AI can help us improve these predictions and batching decisions and help customers get their prescriptions more quickly.” And he said, “If you don’t need a refill for a month, that request can probably wait until we can batch it. If you need a medication urgently, we’re going to bump it up to the front of the line.” So this is how they used all this stuff. Very cool.

And that’s just to give you an idea of kind of what they were thinking about and how they approached it. So here’s what I think about that, and I think the lessons we can learn, generative AI has basically overrun us as a planet, as people, but it’s usually, it really is a lot at still this point about concepts and theories and ideas about what generative AI can actually do. And I felt like the practical use of generative AI, it really is still very nascent. We haven’t seen a ton of proven use cases out there. But I thought that this idea from Amazon and Amazon Pharmacy’s case study is really an important one for the market because it reveals lessons that any enterprise can learn from, and it gives you basically a blueprint that any enterprise can follow in their journey to unlock the value of generative AI. Here’s why I think that. There’s really a few points. There’s three.

So the first one is I wanted to start by saying Amazon culture played a part in this whole scenario. And what I mean by that is if you talk to Amazon, folks that work for Amazon, they’ll always tell you, one thing was they tell you at AWS, they’ll say, “AWS was born out of Amazon’s desire to explore how to better host Amazon Marketplace.” And so these elements of Amazon culture play into this case study. So number one, Amazon typically approaches innovation and opportunities because a customer presents it with a challenge. They’ll always tell you that at AWS and say, “Well, why are you looking at this?” “Well, because the customer asked us to.” So in this case, the customer, Amazon’s many divisions are frequently the customer.

And Amazon Pharmacy has this, obviously this natural connection and path to experimenting with AI experts and the platforms at AWS. So that plays into it. So let’s just say that culture of how they approach things is, do we have a problem? Are we going to look at it? Where can we go within? How do we solve this? And that leads into the second point that I think you take away from this, that if you saw in the brief description we had about what this looks like, Amazon starts with problem solving and they also start with expertise. So let’s look at that for a second. If you think about the approach they took, the team started with this process or operation that needed improvement. All four of those examples we talked about were processes or operations that they said, “How do we be faster? How do we be more efficient?” Really was the goal there when you were hearing those issues they were looking at.

So what I think is key is they started with the business or operation problem first, and then they explored whether there was a technology that could help solve it. And I think this is fundamental, it sounds kind of basic, but you’d be amazed. The most critical first step to being successful with generative AI is that you think about the problem first. You don’t say, “We want generative AI and we’re going to figure out what to do with it.” And second, you think about, this was interesting to me. So that’s number one. It’s very fundamental. But I wanted to note something else here that they were able to move swiftly. The pharmacy was. In part because the team leader, this Alves, has AI experience. And not only does he have AI experience, he has it in the areas that are most relevant to Amazon Pharmacy, the supply chain management.

And the companies that are really going to move fast and operationalize AI the earliest are the ones that have been invested. I’ve said this many times, but it’s a fact. The ones that have been invested in AI for at least two to three years are going to be the ones that are going to move the quickest. And the reason for that is these companies already understand the lifecycle management of AI, the risks, the need for data. And many companies don’t have the luxury of having experienced AI technologists on staff like pharmacy does with Alves.

So you’re learning as you go, right? You don’t have anybody to lean on with experience. So as a company, you’re learning as you go. And I think that when that’s the case, if you’re a company that doesn’t have bench strength in AI, whether it’s business experience or the technical experience, how do you overcome that? The first thing you do is you leverage vendor partners or consultants who do have that and maybe even domain expertise. So that’s something to think about. I think as we keep moving forward in the generative AI era, we’re going to continue to see companies take a step back and look for a little help, a little expertise, which makes them actually run faster once they do have that. So that’s another lesson.

And I think the final one to think about here is Amazon is harnessing the good parts of LLMs and is avoiding the bad parts of LLMs. So I’m going to give you my favorite quote ever about AI, and it’s from Andrew Ng. Ng? Ng. N-G. Who runs Cloudera. No, I can’t think of the name of the company, DeepMind. He was from DeepMind. And so he says this, he says, “Use LLMs as a reasoning engine to process information rather than using it as a source of memorized information.” So I think that’s really wise. It gets at the crux of what the issues are with LLMs.

And if you note, and I’ll get into this for a second. But if you note in Amazon’s use case, the generative AI was not unleashed in text generation or conversational AI agents. It was rather Amazon channeled the generative AI as Andrew suggests, and it’s that as a reasoning engine to process information. And I think as LLMs continue to evolve, some of their challenges might get solved around hallucination and misinformation and just how they’re a little brittle. They’re not ready to be unleashed on us without a lot of help in that sense. But, and I think those challenges are going to be addressed over time, but in the meantime, can your company take the risks that LLMs present when they are used in this way, if they’re used as a source of memorized information? I don’t think so. So looking at them as an engine to process information is a much safer and a really good way to go, as you could see in these examples that Amazon Pharmacy used.

So that’s it today. I wanted to share with you that use case and just some very clear examples about how to move forward. And kind of I think it’s encouraging to see that kind of work out there. It’s pragmatic and practical. I love that. So that’s our show for today. I want to thank you for joining me here on The AI Moment. Be sure to subscribe and rate and review the podcast on your preferred platform, and we’ll see you next week.

Other Insights from The Futurum Group:

Microsoft Copilot Forecast, Fairly Trained, Google ASPIRE | The AI Moment – Episode 12

Lawsuits and Probes, How OpenAI & Microsoft Are Impacting the Trajectory of AI | The AI Moment – Episode 11

Watermarking & Other Strategies for Licensing AI Training Data & Combating Malicious AI Generated Content | The AI Moment – Episode 10

Author Information

Mark comes to The Futurum Group from Omdia’s Artificial Intelligence practice, where his focus was on natural language and AI use cases.

Previously, Mark worked as a consultant and analyst providing custom and syndicated qualitative market analysis with an emphasis on mobile technology and identifying trends and opportunities for companies like Syniverse and ABI Research. He has been cited by international media outlets including CNBC, The Wall Street Journal, Bloomberg Businessweek, and CNET. Based in Tampa, Florida, Mark is a veteran market research analyst with 25 years of experience interpreting technology business and holds a Bachelor of Science from the University of Florida.


Latest Insights:

HP Q1 2024 Earnings Could Be Hiding a Demand Easter Egg Ahead of the Impending AI-Driven PC Refresh Cycle Reset
Olivier Blanchard, Research Director at The Futurum Group, shares his insights on HP Q1 2024 earnings, which send mixed messages about PC demand ahead of the impending PC segment’s refresh cycle reset, driven by the new AI PCs.
Company Banking on Significant Revenue Growth in 2024 with Upcoming Volvo Launch
Daniel Newman and Keith Kirkpatrick of The Futurum Group cover Luminar’s Q4 2023 and FY 2023 earnings and discuss the challenges and opportunities that lie ahead for the LiDAR provider.
The Futurum Group’s Paul Nashawaty and Camberley Bates share their insights on Pure Storage’s earnings and future outlook.
HPE GreenLake Provides Hybrid Cloud Services as a Service
Alastair Cooke, CTO Advisor at The Futurum Group, shares his insights on how HPE has developed the GreenLake portfolio to deliver a variety of hybrid cloud infrastructure as a service offerings.