Elevate — The Global Commerce Summit 2025 | Kapil Dabi, Google Cloud

Clips
More from Elevate — The Global Commerce Summit 2025

Kapil Dabi

America's Market Lead - Retail & Consumer

Google Cloud

play_circle_outline Definition and importance of agentic AI in retail industry.

play_circle_outline Use cases of agentic AI in customer journey, content creation, and virtual try-ons.

play_circle_outline Technologies and platforms enhancing trust and collaboration between humans and AI.

play_circle_outline Future trends in AI focusing on multi-agent systems and closing the gap with consumers.

Info
Transcript

Kapil Dabi, Google Cloud

Kapil Dabi

America's Market Lead - Retail & Consumer Google Cloud

Kapil Dabi, America's market lead of retail and consumer at Google LLC, joins theCUBE’s Scott Hebner and Patricia Wu at Elevate — The Global Commerce Summit, to discuss the growing impact of agentic AI in the retail sector. The conversation explores how autonomous AI systems are reshaping customer engagement and operational efficiency.

Dabi outlines how agentic AI can execute tasks across the buyer journey, enabling intelligent decision-making at every touchpoint. Hebner and Wu examine the role of multi-agent systems in streamlining workflows and suppo... Read more

explore Keep Exploring

What is agentic AI and how is it different from traditional AI systems? add

What are some examples of how retailers are using AI to enhance customer engagement and improve the pre-purchase experience? add

What is the fundamental thing that retailers leading with AI do to set themselves apart from those just optimizing existing processes? add

What are the key shifts happening in the technology landscape in relation to multi-agent systems and AI integration, and how are retailers adapting to these changes? add

bolt Powered by CUBE AI

Kapil Dabi, Google Cloud

search

Patricia Wu

>> Welcome to Elevate -- The Global Commerce Summit. We're coming to you from the Miami Beach Convention Center. I'm Patricia Wu, your host, along with my co-host and analyst, Scott Hebner. And you're going to see Scott grinning from ear to ear because we're talking agentic AI. Nothing makes you happier.

Scott Hebner

>> I love it. I love it. Yes.

Patricia Wu

>> All right. And we're going to be talking with Kapil Dabi, America's market lead, retail and consumer for Google Cloud. Thank you so much for being here.

Kapil Dabi

>> My pleasure. Thank you for having me.

Patricia Wu

>> So, I'm going to toss out the first question because I know he's chomping at the bit. He's got to have a lot more.

Kapil Dabi

>> Yeah, sure.

Patricia Wu

>> Talk to us about agentic AI in the context of retail and just how pivotal is this shift from what we've seen before?

Kapil Dabi

>> I think that's a great question and a lot of retailers actually do ask us, a lot of our customers do ask us, "How real is agentic AI?" or, "what is agentic AI?" But let's unpack this thing a little bit. Agentic AI back in the days when we think about AI, it was about recommendation. So, you ask a question, it'll come back with a summary or some recommendation. Agentic AI is taking that into the next level where it can actually take autonomous decisions, which it means that it can actually learn, adopt, and execute task on your behalf. And that's the fundamental shift which we are seeing in the industry. And the other thing is, the easier way to understand this thing is think about if you break it out. And for easiness, I always think about a store associate is doing some tasks. Can an agentic AI can actually help execute some of those tasks? Or when you think about customer service, can a customer service rep now have all the summarization, all the micro-moments when a customer actually interacts? All those things are actually already being given to the customer care rep, so when actually a customer calls them, they already know what is he calling about? So, agentic AI is basically executing actions on your behalf, learning, adapting. And then, it's all about driving things in an autonomous fashion and that's basically the ultimate culmination. We are also talking about not one single agent, which can actually perform a task. We are talking about multi-agent systems. So, think about a complex workflow like supply chain. So, you might have a separate agent for assortment, separate agent for fulfillment or replenishment. How do those agents talk together and can actually execute on your behalf, without human intervention or with even human intervention is also fine? The goal is basically empowering humans. It's amplification of how you can actually do the collaboration between humans and AI. The goal is basically how do you improve customer experience? So, agentic AI, from a retail perspective, is all about eventually business outcomes, which is basically, can I drive my customer experience in a much better way? Or can I improve my top line? Or can I actually improve my bottom line? Just to give you an example.

Patricia Wu

>> So, what are some of the most compelling ways that you're seeing retailers use this agentic AI?

Kapil Dabi

>> And I often talk to most of the retailers and our customers in a very simplistic fashion to just break it down, the key thing is when you look at a customer cycle or customer journey from path to purchase, it is about, in very simplistic term, it's pre-purchase, purchase and post-purchase. So, if you think about that customer journey, it's very complex when you really do this, but for simplicity, let's talk about that. So, when you think about pre-purchase, think about customer engagement. Think about am I getting the right promotion? Is the image which has been created in my promotional element, does it resonate with me? So, when I'm looking for a blue jacket and I bought this thing from my past purchase, should it give me a matching recommendation out of it? So, retailers are using agentic AI to actually do content creation, which means that they're currently creating images and videos of the products using generative AI. You're no longer clicking pictures, you're actually doing that. We have retailers who are actually creating multimodal ads. So, we have clients which actually operate in multiple countries. You create a single ad and you can do translation, you can do localization with generative AI. So, today you can create an ad in English, but you can use the same ad in French or in Spanish. So, that's basically the engagement piece. During the purchase, we are also seeing multimodal search. Now, you can take a picture of shoes, which you're wearing, and then you can do the search from there. AI beauty advisors have been there. So, when people are buying beauty products, you can actually take a picture of it, you can select lip balm, foundation, mascara. And it'll do the product discovery for you across multiple brands, and then basically we'll do the recommendation. But after recommendation you can say, "Please purchase it for me." It'll also purchase it for you. We are also seeing conversation AI where you can actually talk and in the context, so you can actually say, "Hey, I'm actually in Miami. I'm going to a gala party. The theme is blue. Can you please recommend me certain products?" And it's actually going to recommend you products and you can keep on talking to it. It's like, "Okay, that's great. These are all great blue dresses, but I want sleeveless because it's too hot and humid. Can you suggest because I'm going to be on a boat?" So, now you're actually maintaining the context and the intent of the purchase. So, these are some of the great examples of how agentic AI and gen AI is actually helping the end-to-end retail value chain.

Patricia Wu

>> Wow, that sounds like a dream come true for me. I hate being in the stores and I know what I want.

Kapil Dabi

>> Yes.

Scott Hebner

>> Yeah, and I think what you heard was that's a lot of what generative AI can do, right?

Kapil Dabi

>> Correct.

Scott Hebner

>> It helps you analyze information, it can create content, it can automate simple tasks, things of that nature. It's not designed, generative AI though, to be able to make actual decisions and to understand goals, help you problem solve, chart your path forward to accomplishing whatever you want to accomplish. So, you got to take a agentic AI core and surround it with additional technologies. Can you just talk a little bit about what those technologies are and how you see them coming together? Because when agents get to a point where they can actually make trustworthy decisions and actually tell you why one decision may be better than another one and how to go out doing it, it gets really powerful.

Kapil Dabi

>> Yeah, it means most of the technologies powered, especially from a Google perspective, we have a lot of models which we have generated and invested like. So, Gemini 2.5 Pro can actually do a lot of decisions. We have also launched deep research agents, which can only focus on deep research. We have also have agents, which we have recently launched called idea-generation agents. So, a lot of these models, we have models for video creation, image creation. Actually, at Google Next we also launched, now you can do prompt to music, so you can actually create your own music. So, the idea here is the technology is all infused for us is through these new models, which we have created. And there's a reason why there's no one model for all. And there's a reason because the model itself, the LLMs are trained in such a way that they're best to do that specific role. So, when we think about Veo 2 and yesterday we actually launched Veo 3, which is basically our model to create videos with texts, it is optimized to the creation of the videos. So, the models are trained in such a way that it understands that the videos needs to be created. So, that's one part. But one of the fundamental shift with the way Google Cloud is leading to help retailers is we have a platform approach. So, the idea is that models tomorrow is going to become a commodity, but when you're looking at creation of business impact at scale, when you're thinking about AI at scale, the most important thing you have to do is think about a platform approach. So, we have platform which has the core infrastructure layer to database layer to a development platform for our developers. But then, you have agents can be built on top and then those agents can basically also orchestrate with third-party agents and so on and so forth. So, it's truly a platform play. And the reason why always we lead with a platform play is tomorrow, you as a retailer might look at a different model than you might use to use Llama or you might use to use Gemini and other use cases. The goal is that whatever is sufficient or whatever is adequate for your business case, you should be able to use. And that is the power from a model perspective, Google is actually giving a lot of retailers. So, it's giving the power to the retailer to make a call of how you would like to run your business and however you want to impact your business, but you have all the tools and technologies to actually enable you.

Scott Hebner

>> And in the world of business, and particularly in commerce, trust becomes the currency of innovation.

Kapil Dabi

>> Yes. Yes.

Scott Hebner

>> If you don't have trust, you're not going to get any ROI. Trust is a function of people understanding why the AI is telling me to do something, being able to explain it. So, I'm curious to what Google's view is of enhancing trust. And I don't know if you've saw the New York Times article a few weeks ago that basically talked about as these generative AI models get more and more sophisticated, the hallucination rates and the bias is going up versus down, and a lot of people see that, right? So, what do you think the future holds in terms of really making these trustworthy?

Kapil Dabi

>> Yeah, it means Google being a market leader, we have a responsible AI policies means those are public policies. You guys can go read on those policies. But it is all about at the end of the day, being transparent, being fair, being trustworthy. One question often gets asked, "Are LLMs going to be trained on my data?" And your data is your data. So, retailers or enterprise, whoever is using Google data is we don't take that data. We don't train LLMs on that data for others. Your data is your data. And anytime when there is an application or use case which we would want to implement, like this content creation. So, if you want to do content creation for yourself, it is grounded in your enterprise data, which means that the brand guidelines is followed, which is your brand's guideline. We are not going to create an image or a video outside that. So, the responsible AI is actually the crux of everything what Google does.

Scott Hebner

>> And coming off of Google Next and Google I/O what was the big news there?

Kapil Dabi

>> Yeah, there are many, given that it cuts across Google, but for what excites me the most, actually, it's Google Next, was the launch of our Agentspace. Agentspace was launched earlier, but there are a couple of enhancements which we had announced, like agent-to-agent protocols. So, when I talked about that multi-agent system, how do you solve a complex workflow like supply chain or a product design? The whole concept is that multiple agents needs to talk to each other, whether it's your first-party data, agents or third-party agents. And agent-to-agent protocol was announced. Another big announcement at Google Next was Agent Developer Kit, ADK, which means it's very easy to now build, configure, and deploy your own agents, right? Google I/O just happened yesterday and there was such fantastic announcements, but what excites me the most from a retail perspective is there was a launch of AI mode, which means that today when you do Google search, there is an AI mode. It actually enhances your search. It'll also remember your past searches. It can take certain actions on your behalf. Like, "Hey, I'm going to Miami for two days. Can you suggest hotels and can you book hotels for me?" We also launched virtual try-ons. Now, you're wearing a blue dress. Now, you can actually use the virtual try-ons and you can actually, in real time, you can flip the pictures and it'll actually help you do the try-ons. We also launched agentic checkout, which means that while you're doing this shopping, you can easily say, "I want to shop this thing later," but you can define your price point. You can define your quality shoe size, some certain product attributes, and it'll actually buy it for you, it will do the checkout and all that stuff. So, these are all real agentic AI actually helping you shop. So, these were some of the cool announcements, which we just had at Google Next and Google I/O.

Patricia Wu

>> So,. Before I let you go, this sounds like an amazing new world.

Kapil Dabi

>> Yes.

Patricia Wu

>> What sets apart the retailers that are leading with AI and reimagining all of this from the ones who are just simply optimizing existing processes?

Kapil Dabi

>> Yeah, that's a great question. The fundamental thing is what can you do to... Eventually, it's all about collaboration or identifying opportunities and landscape for human and AI to collaborate together. I think that's the fundamental thing which every retailer should be thinking. So, there will be always tools and technologies, but I think fundamental thing is what do you stand for? What are your brand values? And you talked about trust, loyalty, creativity. There's so many things that brands actually stand for, retailers stand for. And I think they're all, at the end of the day, powered by humans. So, the goal of AI is to actually empower humans, it's to amplify what humans are doing, it's to unlock their creativity, it's to unlock relentless innovation they might potentially do today with AI. So, at the end of the day, it is the unlock of human and AI potential together, and that's what I truly believe that retailers really need to lean upon. One thing I always tell us earlier, the narrative used to be what is your AI strategy? I think that narrative has shifted. The narrative has shifted is like what is your business strategy and how can AI actually enable you to achieve that business strategy? So, in essence, what we are saying is AI is there to actually help you achieve what you're looking to achieve, whether it's trust, loyalty, creativity, launching of new products, whatever it is for your retailer or for a brand to come and achieve those things. It is actually, in essence, finding the right collaboration between human and ai. And that, in essence, is the key.

Patricia Wu

>> So, I was just going to say the AI is no longer separate? It is embedded, inherent?

Kapil Dabi

>> It is embedded inherent in everything you are doing, and that's where you're actually going to talk about getting the true value and true, I would call it, how retail is actually winning in the new stages. How do you unlock the potential of AI with humans together? And that's the winning, I would say mantra.

Scott Hebner

>> So, our last question before we have to wrap here, but if we were here a couple years ago, I think we would've been talking about generative AI.

Kapil Dabi

>> Yes.

Scott Hebner

>> We're talking now about agentic AI.

Kapil Dabi

>> Yes.

Scott Hebner

>> So, in two years, we're here again. What do you think we're going to be talking about?

Kapil Dabi

>> It's hard to say because the change is actually happening at a rapid pace. I would say we are talking about multi-agents and multi-agent system from a technology perspective. So, I think I'll give you two sets of answers. One is basically capability perspective and the capability perspective, I think we are talking about multi-agents, okay? Multi-agents is not there today. We're talking about for the future. So, in 12 to 18 months, I think we are going to see complex workflows, like product design in retail. Like I was a practitioner and operator in my past life from the industry, very difficult to basically compress your product design from six months to two months. It's not possible, until you have, can you do sketch to image with generative AI? Can you do product design with generative AI and so forth? So, I think we are talking about some of these ideas. The question is how do you orchestrate multi-agent systems? How do you truly take a business problem and compress the timelines or whatever your KPIs are? I think that's where the shift is going to happen. The other shift from, again, from a change management retail perspective is I would say the consumers are far ahead from an AI perspective. We do Google search and there's AI overviews in our AI mode. We do Instagram and Meta is there and we do WhatsApp and AI's infused in everything, what we are doing. I think that the retailers really need to step up with the change of technology landscape they have to leapfrog. So, dichotomy today is consumer is far ahead retailers, some retailers ahead and some retailers are behind. So, how do you fill that gap? So, that I think is the important shift which you're going to see is some retailers going to really leapfrog, going to use the technology capability, fill that gap and the void where the consumers are. And then, some really have to then figure out where they perform best. So, those are the two perspectives, multi-agent system from a tech capability perspective, but then also from a change-management perspective, how do you fill that void and the dichotomy between consumer and the retailers?

Scott Hebner

>> I think that's right on. That's what I've been hearing.

Patricia Wu

>> I have a feeling that we'll be hearing a lot more from you and Google Cloud about innovation to fill that gap.

Kapil Dabi

>> Yes, absolutely. It's my pleasure to be here. Thank you so much for inviting me.

Scott Hebner

>> Thank you.

Patricia Wu

>> Thank you so much.

Kapil Dabi

>> Thank you.

Patricia Wu

>> Well, we have a lot more coming up from Elevate -- The Global Commerce Summit, so don't go away. I'm Patricia Wu for Scott Hebner, and we will be right back. You're watching theCUBE, the leader in enterprise tech news and analysis.