How Agentic AI Will Transform Retail Forever With David Dorf Of AWS | 5IM

Summary

Omni Talk Retail Podcast Summary

Title: How Agentic AI Will Transform Retail Forever With David Dorf Of AWS | 5IM
Host: Omni Talk Retail
Guest: David Dorf, Head of Retail Industry Solutions at AWS
Release Date: April 6, 2025

---

Introduction

In the April 6, 2025 episode of Omni Talk Retail, hosts Chris Walton and Anne Mezzenga delve into the evolving landscape of artificial intelligence (AI) in the retail sector. The episode features an insightful conversation with David Dorf, Head of Retail Industry Solutions at Amazon Web Services (AWS), focusing on the transformative potential of agentic AI in retail. The discussion highlights the distinctions between generative AI and agentic AI, current applications within the retail industry, future implications, and strategic recommendations for retailers looking to leverage these advanced technologies.

---

Defining Agentic AI

David Dorf begins by distinguishing agentic AI from the broader category of generative AI (Gen AI). While Gen AI is primarily concerned with creating content, agentic AI focuses on taking autonomous actions.

David Dorf [00:31]: “Genai is about creating content, while agents are really about taking action. They both use foundation models underneath, but there's really three key things for agents...”

Key Characteristics of Agentic AI:

1. Autonomy: Agents operate with minimal human oversight, executing tasks independently.
2. Reasoning: They utilize complex reasoning processes, breaking down problems into manageable steps, often referred to as a "chain of thought."
3. Tool Integration: Agents have access to various tools and data, enabling them to perform specific actions based on their objectives.

Dorf emphasizes that unlike Gen AI, which might generate an image upon request, agentic AI can undertake multi-step processes such as creating the image, posting it to a website, and ensuring it meets specific criteria.

---

Current Applications of Agentic AI in Retail

When asked about the current utilization of agentic AI in the retail industry, Dorf provides several compelling examples, illustrating the early adoption and significant impact of these technologies.

David Dorf [01:37]: “A great example is Amazon uses agents to do a lot of Java upgrades... saved us about $260 million last year.”

Notable Use Cases:

1. Operational Efficiency:

   • Amazon: Utilizes agentic AI to manage and upgrade approximately 30,000 Java applications, resulting in cost savings of around $260 million annually. These agents not only perform upgrades but also build unit tests and maintain documentation.

2. Data Management and Analysis:

   • Tapestry (Luxury Retailer): Employs an agent that assists in data retrieval and query responses. Users can input natural language questions, which the agent translates into SQL queries to extract data and provide comprehensible answers, functioning similarly to a human analyst.

3. Customer Recommendations:

   • Tire Retailer Example: Implements agents that provide personalized tire recommendations based on vehicle make and model, usage patterns, and other relevant factors. These agents also explain the reasoning behind each recommendation, enhancing customer trust and decision-making.

---

The Future Impact of Agentic AI on Retail

As the conversation progresses, Dorf explores the potential future transformations agentic AI could bring to the retail sector over the next five years.

David Dorf [03:24]: “This is where things can get a little bit crazy... how do I affect a person when they're using an agent?”

Potential Transformations:

1. Automated Shopping:

   • Computer Use Agents: Announced by Anthropic and OpenAI, these agents can control browsers, navigate websites, and perform purchases autonomously. An example highlighted is an OpenAI demonstration where an agent used Instacart to purchase groceries, including a news feature in The New York Times where an author bought eggs via an agent.

2. User Interface Optimization:

   • Current retail websites are designed for human interaction. With the rise of agentic AI, there may be a shift towards agent-optimized sites, similar to the mobile optimization trend, to facilitate easier interactions between agents and online platforms.

3. Advertising and Marketing:

   • Traditional advertising strategies may become less effective if consumers use agents to handle purchases. Dorf raises critical questions about the future of search engine optimization (SEO) and personalized advertising, proposing that retailers might need to focus more on promotions and loyalty programs that influence the end-user rather than the agents.

4. Strategic Adjustments:

   • Retailers need to think strategically about how agentic AI will change consumer behavior and interactions, potentially reimagining aspects like loyalty programs, promotional tactics, and overall customer engagement.

---

Strategic Recommendations for Retailers

In addressing how retailers can begin to harness the power of agentic AI, Dorf provides pragmatic advice focused on both immediate and long-term strategies.

David Dorf [05:42]: “Number one is basic blocking and tackling... Number two, there's a lot of basic gen AI use cases...”

Recommended Steps:

1. Foundational Practices:

   • Avoid getting swept up in the hype surrounding Gen AI. Instead, focus on implementing proven, high-impact use cases that offer clear returns on investment.

2. Immediate Applications:

   • Product Descriptions: Utilize Gen AI to create and optimize product descriptions, enhancing both SEO and customer engagement.
   • Shopping Assistants: Deploy AI-driven shopping assistants (e.g., Rufus) to improve the customer shopping experience and increase sales conversions.

3. Exploring Agentic AI:

   • Begin integrating agentic AI into business processes to automate tasks and enhance efficiency. Rather than replacing human workers, these agents can augment human capabilities, allowing staff to focus on more complex and creative aspects of their roles.

4. Preparation for Future Changes:

   • Proactively consider how agentic AI will reshape various facets of retail, from website design to marketing strategies, and develop plans to adapt accordingly.

---

Conclusion

David Dorf's insights underscore the pivotal role agentic AI is poised to play in the retail industry's future. From enhancing operational efficiencies and personalizing customer experiences to fundamentally altering how consumers interact with retail platforms, agentic AI presents both significant opportunities and challenges. Retailers are encouraged to adopt a balanced approach, leveraging foundational AI applications while strategically exploring the emerging capabilities of agentic AI to stay competitive in an evolving marketplace.

Chris Walton [05:16]: “Well, David, you just blown both of our minds here...”

The episode concludes with an acknowledgment of the groundbreaking potential of agentic AI, leaving retailers with actionable strategies to navigate this transformative technological landscape.

---

Remember: As agentic AI continues to evolve, staying informed and adaptable will be key for retailers to harness its full potential and maintain a competitive edge in the industry.

Transcript

[A]

Foreigning us now for five insightful minutes is David Dorff. David is the head of Retail industry solutions at AWS. David, welcome back to Omnitalk. And let's get started with this. Gen AI was the big topic for the last two years, and now 2025 seems to be the year of the agent, or agentic AI as some might call it. How do you define, quote, unquote, agents?

[B]

Ah, yes, good question. So genai is about creating content, while agents are really about taking action. They both use foundation models underneath, but there's really three key things for agents. Number one, they're autonomous, so they have a role and they require minimal human oversight. Number two, there's reasoning involved, so they actually do this thing called chain of thought where they break down problems into smaller steps. And number three is they typically have some sort of tools or access to data in tools to be able to take action to do something. So those three things are really what set agents apart. And for example, you can use generative AI to just create an image, but an agent, you could say, I need a specific image that is relaxing and I want it posted to a website. And it can do all of that. Right. It's taking multiple steps and it's doing action. So there are lots of different genai use cases out there, but agentic genai is really where things are starting to explode.

[C]

So, David, are retailers actually using these agents already today? And if so, how are they using them?

[B]

Yeah, it's not quite as prevalent right now because it is pretty new, but for sure. So a great example is Amazon uses agents to do a lot of Java upgrades. Sounds like a boring task, but we have like 30,000 Java applications that need to be upgraded from an older version to a newer version. And we have agents that go through and do that and saved us about $260 million last year. Wow. And in addition to just doing those upgrades, they'll also build unit tests and they'll do documentation and things like that. And we've got some other retailers doing some cool stuff like Tapestry, the luxury retailer. They actually have an agent that helps find data and answer questions. It will take your textual question, figure out what database it needs to answer that, convert it to SQL, do the SQL query, and come back with it with an English answer. So it's kind of like an analyst, but it gives the business a chance to be able to get this information without necessarily having to create all sorts of weird SQL queries. And then we have a tire retailer who's giving recommendations for tires. So You. You, for example, can say what make and model you have and what you use your car for. And it goes and hits several different resources to figure out what the best tire recommendation might be. And it kind of gives you the reasoning behind why it recommends one tire over another, which I think is pretty unique.

[A]

Wow. And I don't know about you, but I had no idea the agents were already being deployed to the degree which David was saying. I thought this was just kind of the new buzzword idea for 2025, but. So with that said, David, I'm curious. How do you see agents changing retail even more over the next, like, five years? Let's say, what are the big things that are still going to change on this front?

[B]

This is where things can get a little bit crazy. One type of agent is a computer use agent. So those were announced by Anthropic and OpenAI. And basically it has control of your browser and so it takes a picture of the browser screen so it knows where things are and it can navigate with the mouse and keyboard. And the. The demonstration that OpenAI gave was using Instacart to do your grocery shopping. There was even a New York Times article where the author bought eggs using a computer agent, just saying, go find me eggs. So if you think about that for a minute, how does that really upend the retail industry with shopping? If people are using agents to do the shopping for them now, they're not going to do it for, like, fashion, but for replenishable items, grocery, that sort of thing. I can see that happening today. The user interface is made for humans and it's cluttered. And maybe we need to optimize that for agents just like we did for mobile. We have mobile optimized sites now. Maybe we'll have agent optimized sites that make it easier for agents to buy things on people's behalf. And then you have to start to think about what are some of the other things that could happen? What happens with advertising? There's no use in advertising to an agent. Should I be personalizing? Maybe that doesn't make sense. How do I affect a person when they're using an agent? Through things like promotions. An agent doesn't care about a promotion. So how do I get that information to a person? What about loyalty? That kind of gets thrown up if I can't really affect behavior. And maybe something like even search. We have a whole industry around search engine optimization. What if it's agents that are out there doing the shopping for us? How is search engine optimization going to change and how do we influence things differently? I think we're on the cusp of some really interesting changes to the retail industry that agents could bring us and we need to start thinking strategically about what are the ramifications here and how can we prepare.

[C]

Well, David, you just blown both of our minds here. Unfortunately we have to get you out of here on this. But we will be thinking of many more questions. Where do people start? There's a lot that's changing. What do you recommend that retailers do now if they're wanting to explore taking advantage of gentic AI or just to kind of tackle some of the things that are going to be happening in the future?

[B]

Yeah, yeah. So number one is basic blocking and tackling. Right. So don't get too caught up in gen hype. There's a lot of it out there. Number two, there's a lot of basic gen AI use cases that you can get a real good return on investment in things like creating product descriptions, definitely being able to do things like shopping assistants like Rufus. Those are some low hanging fruit that retailers should look at. And then we can start to get into the agentic areas where you've got processes that can be automated and you can really start to, I don't, I don't want to say replace people because we're not ready to do that yet, but really enhance the things that people are doing to be more efficient.

[A]

Great stuff, David. Thank you.