In this interview from Google Cloud Next 2026, Karthik Narain, chief product and business officer of Google Cloud, joins theCUBE's John Furrier and co-host Alison Kosik to discuss how Google is architecting the full-stack agentic enterprise. Narain traces the company's investment in purpose-built AI infrastructure back to 2015, explaining that intentional TPU development — long before it was fashionable — was designed to meet the eventual scale demands of AI. He outlines three layers of enterprise adoption: individual productivity, cross-functional process transformation and R&D acceleration, citing Reliance Retail's 7% increase in basket size as a concrete example of what outcome-oriented agentic workflows can deliver to the bottom line.
The conversation also explores how Google is removing barriers to agent adoption through open standards and a three-part agent strategy. Narain details the open-sourcing of the Agent Development Kit (ADK), the Agent-to-Agent (A2A) protocol and the Agent Payment Protocol (AP2), alongside embracing Apache Iceberg and the Model Context Protocol. He addresses vendor lock-in concerns directly, noting that Vertex AI already serves as the inference platform for Anthropic's models and a wide range of open source alternatives, and that Google's knowledge catalog now federates data across AWS, Azure, Salesforce and SAP. Narain also describes ISV partnerships with Thoma Bravo and Vista Equity, whose 350 combined portfolio companies will adopt Gemini Enterprise as their shared agentic platform. From developing an audio-to-audio model in direct response to call center quality degradation to treating continuous model evolution as forward compatibility rather than disruption, he offers a candid roadmap for how Google Cloud is turning product and go-to-market strategy into a single, tightly coupled engine.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
Google Cloud Next 2026. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Register for Google Cloud Next 2026
Please fill out the information below. You will receive an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for Google Cloud Next 2026.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
Google Cloud Next 2026. If you don’t think you received an email check your
spam folder.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Sign in to gain access to Google Cloud Next 2026
Please sign in with LinkedIn to continue to Google Cloud Next 2026. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Karthik Narain, Google Cloud
Karthik Narain of Google Cloud, chief product and business officer, appears at Google Cloud Next '26 to discuss Gemini Enterprise, agentic Data Cloud and Google's full-stack artificial intelligence strategy. Narain outlines Gemini Enterprise as an end-to-end system that spans intelligence to action and is supported by agentic Data Cloud and agentic Defense; they emphasize the role of interoperable agent standards such as the agent development kit ADK, agent-to-agent A2A and agent-to-platform AP2.
Hosts Alison Kosik and John Furrier guide the conversation, probing product strategy infrastructure such as tensor processing units TPUs and Vertex AI open standards and partnerships that aim to enable autonomous context-aware agents across enterprises. Narain stresses a full-stack approach that integrates models purpose-built infrastructure and enterprise data to accelerate adoption, and highlights the importance of outcome-driven process transformation individual productivity gains and research and development acceleration. Partnerships with Thoma Bravo Vista Equity and Reliance Retail and open tooling in Vertex AI underscore Google's multi-cloud and model choice stance.
In this interview from Google Cloud Next 2026, Karthik Narain, chief product and business officer of Google Cloud, joins theCUBE's John Furrier and co-host Alison Kosik to discuss how Google is architecting the full-stack agentic enterprise. Narain traces the company's investment in purpose-built AI infrastructure back to 2015, explaining that intentional TPU development — long before it was fashionable — was designed to meet the eventual scale demands of AI. He outlines three layers of enterprise adoption: individual productivity, cross-functional process tr...Read more
Karthik Narain
Chief Product and Business OfficerGoogle Cloud
In this interview from Google Cloud Next 2026, Karthik Narain, chief product and business officer of Google Cloud, joins theCUBE's John Furrier and co-host Alison Kosik to discuss how Google is architecting the full-stack agentic enterprise. Narain traces the company's investment in purpose-built AI infrastructure back to 2015, explaining that intentional TPU development — long before it was fashionable — was designed to meet the eventual scale demands of AI. He outlines three layers of enterprise adoption: individual productivity, cross-functional process tr...Read more
exploreKeep Exploring
What is Google's vision for the agentic enterprise?add
What is the strategy for developing, standardizing, and deploying AI agents — including interoperability, payments, partnerships with ISVs/SaaS, and enterprise agent platforms?add
Will Google make the technology behind Gemini and its full‑stack capabilities available to external customers—allowing them to run and fine‑tune their own (including open‑source) models on Google Cloud, use Vertex AI for serving, and integrate data across multiple clouds and third‑party platforms?add
How do customer feedback and real-world use cases influence product development and changes in sales approach (for example, moving from selling SKUs to delivering integrated solutions)?add
What are the three top priorities you are currently working on?add
>> Welcome back to Google Cloud Next '26. We are streaming live right here in Las Vegas. I'm Alison Kosik alongside John Furrier, and we're about to get into the future of agentic Enterprise. I want to bring in Karthik Narain. He's the chief product and business officer of Google Cloud. Welcome to theCUBE.
Karthik Narain
>> Thank you for having me here.
Alison Kosik
>> So let's get into it. What is Google's vision of the agentic Enterprise?
Karthik Narain
>> I think the vision of the agentic Enterprise is that we build a system that creates a connective tissue between people, data and the business outcomes of enterprise. And to do that, we want to create the entire stack starting from Gemini Enterprise, which from being just the front door to Intelligence is now an end-to-end system from intelligence to action. We're bringing the agentic Data Cloud to become the context layer for all these agents to operate autonomously at scale. We're bringing agentic Defense so that all these activities can be happening in a protected environment. And we are powering that with the full stack that everybody is talking about.
John Furrier
>> Karthik, I got to say, I'm very impressed. I want to ask you, when did the light bulb go off or when was the moment you said, "This is looking really good." The products are looking good. Gemini, front and center, front door, but the data cloud really powering everything and you got the infrastructure piece. Everything's kind of popping in the stack. You got the control plane, orchestration with Gemini Enterprise, as well as the user experience. When was the moment that you all came together, could you just share some insight into the product and the intention behind this?
Karthik Narain
>> Yeah. I think we've been talking about this full stack in the last year or so. Everybody's talking about it, but Google has been intentional in building this full stack since 2015, starting with TPUs. When there was a lot of noise in the industry that Google sometimes tries to be cute by wanting to build everything in house, but the intention was, we knew the scale of AI and you need purpose-built infrastructure to do that. Two or three years ago, when we went deep into converting our transformer experience into building Gemini, it was very clear that there were three factors that was important to build a state-of-the-art or SOTA model. The three factors were you need to have the world's best AI research team. You need to have original data on which you could train, and you need to have the latest and greatest infrastructure to be able to do that. When we were moving into the world of inference, we started thinking about what are the three things that makes sense for organization to increase adoption? That is high quality reasoning that is grounded on enterprise data. The second aspect is real-time intelligence with low latency and high performance. And the third thing is scalability through predictable and efficient cost. When we thought about it that way, we thought that this is an opportunity to create that full stack that we give organizations and help them focus on converting their deterministic linear workflows to agentic AI first workflows and put the onus on enterprises to focus on that while we give the instrumentation that is required to enable their employees, use that, and a platform with which they can scale across the enterprise.
John Furrier
>> Yeah. And I got to say the timing is impeccable because the enterprise adoption has not been as strong as some people have predicted. This year, we are predicting that's going to explode pretty big. Now you saw rag, you saw marketing copy, first generation of AI would work well, but coding really opens the door. Once coding happened, you started to see the enterprises leaning into the coding, obvious productivity gains. You're seeing stats where the coding internally is outpacing humans, which is good. Now agents are coming right behind us. So talk about that dynamic for the enterprise because now they've got coding, now they got agents coming on. What are the critical factors and requirements they need to have in place in the full stack that you're seeing with customers to really turbocharge the agent proliferation?
Karthik Narain
>> Yeah. We look at enterprise adoption to happen in three layers of an enterprise. The first one is enabling the people, their individual productivity to get intelligence and information cross reference through the organization that they otherwise need to be talking to 50 different people to put things together over a month's time. For example, if you want to launch a promotion of a product in a particular region or determine price for a product, you could feed all the information from product to CRM to sales information to competitor information and ask for this and get this marketing campaign created in 30 minutes. So that is the individual productivity, including coding capabilities and instrumenting that. The second layer is where the biggest bank for the buck is going to happen, which is process transformation. In the traditional world, every process was siloed to either a department or based on how an application stack was built to do that. If there was a CRM that was just a sales cloud, that only focused on sales, marketing was focused on by a different product and so on and so forth. But when you want to, again, think about winning a customer, you need to be thinking cross-functional. So this is the opportunity to build those agents to be outcome oriented and give it the data to do that. So that's what we are thinking and we are helping, and there are several examples. Yesterday we talked to Andy, say from Citi where the Citi Sky product is not just going to change wealth management for Citi, but he believes, he's a seasoned executive and I could see the energy and the excitement in his eyes that it's going to change wealth management and scale it. So that's the process layer. And the third element is R&D. And this is going to happen in healthcare, this is going to happen in material sciences, and it's going to happen in three dimensions. It's not just about creating the next breakthrough product. It's also about how do I create the product that I was always thinking about creating, but I did not have the budget to create it. Now I could do it in half the cost, or this product would take otherwise two years to create it and I would miss the market opportunity. Now I can launch the product in six months. So that's what we are seeing. And these are examples of what customers are doing.
John Furrier
>> Are you guys priming the pump? I noticed you guys are providing agents. Google has agents. I can build my own agents. So there's a lot of kind of... I don't want to say acceleration going on inside the customer. Take us through that piece because the adoption curve, the demand is high, but the speed of change is so fast. Some folks get paralyzed by that. They're like, "Whoa, I just learned. I just locked in. Now something new and better's here." So this is a real innovator's dilemma because they're jumping in. How do you guide them through that process? Is it blueprints? Is it agents? Because they recognize they got to lean in fast.
Karthik Narain
>> Yeah. We're doing it in three ways. Number one, we have created the standard for the industry to develop agents so that those agents are interoperable. We developed the agent development kit ADK and we open-sourced it. We developed the A2A protocol, open-sourced it. When commerce and real business needs to happen, you need payments. So the AP2 agent payment protocol we created and we are also, it's not invented here syndrome. We are also embracing open standards, whether that's Apache Iceberg or model context protocol. So that's one way we are doing. The second is the agent strategy is going to be, again, three part. One is we are creating a lot of first party agents. Yesterday we launched the deep research agent for intelligence or insights. And there's going to be several agents that is going to be made by Google. The second thing is we are partnering with ISV SaaS companies. In the last two weeks, we announced a deep integration partnership with Thoma Bravo and Vista Equity. These are the world's two largest ISV or SaaS private equity. The arrangement is all their portcos are going to underlying their software. They're going to innovate on their core capability, but they're going to use Gemini Enterprise as the agentic platform, which means that tomorrow you build an agent in Anaplan or Coupa or Duck Creek, those agents will show up in Gemini Enterprise. And from Gemini Enterprise, you can directly create a Coupa agent and so on and so forth. So that's the second part strategy on agents. And the third part is we created a complete agent platform for enterprises to build their own organization specific agents that others can do that. So that's the agent story. And the third is a platform approach. And that's why we've created the end-to-end life cycle starting from giving agent a cryptographic ID for identity, to registry to gateway to security all through the lifecycle.
John Furrier
>> It's a great point about open source. I'm really glad you brought that up. And you guys are very intentional too, by the way. You're giving some candy to the kids and the developers out there. But the private equity angle is interesting because the number one question we get on theCUBE in kind of C-suite is, the question is, if we could redo our business today from scratch, start over, how would we do it with today's tools? So obviously the acceleration, I can see them kind of repivoting. That's a question that comes up. How would you answer that question? The CEO says to the team, "Hey, let's do a shadow company, not shadow AI or shadow IT. What would we do if we started the company over today?"
What would it look like? It wouldn't be a bolt-on. It would be a reinvention. What would you say?
Karthik Narain
>> Yeah, I think usually that's a provocative question that a customer asks and they generally will not have patience for more than a day to walk through that path of recreating the company from scratch. That's just a provocative question. They basically are saying, "How do we not just do small changes? How do we do a big change?" But they also want to leverage all the investments that they've already made because there's going to be this question of who's going to fund this entire transformation and what is the time it's going to take for the transformation, even if it's worth it. Am I going to be busy replatforming my company? So the way we approach this is, the big change that they need to do is identify processes where they put outcome first, which means that if you are thinking about the customer sales journey, your outcome should be that, how do I convert the intent to a product delivered home? And for that, what are all the things that I need? They may start learning about the products through social platforms or through Google search surfaces and so on and so forth, or Gemini surfaces. And from there, they would want to come to their website, and from there, there should be a seamless orchestration of their intent into the product catalog that shows, and there should be a seamless way to add on the additional products that they are able to do and they should take it forward. For example, Reliance Retail did this and they used Gemini across their platform. And as a result, they're actually able to increase the basket size by 7%, right? This is a big deal. The 7% revenue they did not have.
John Furrier
>> That's revenue.
Karthik Narain
>> And it's the same amount of time that the same customer was spending instead of them checking out for $100 or 100 rupees, it is $107, right? Just imagine every customer going. So reimagine that, but by leveraging your existing platform and using the agent as the intelligence orchestrator, and the model's role is going to be to create continuity of context from these influencer surfaces to your merchant page, to your catalog, to the payments, and the incentives that comes with payments.
John Furrier
>> I think what you're saying is right. This is what the focus is. It's not just cost reduction. It's revenue. People are starting at the structural foundations where the revenue is generated and go there.
Karthik Narain
>> Totally.
John Furrier
>> Why not? Go there first. All right. I have to ask this question. I was talking to some press folks and some analysts and some customers. And because Gemini is so featured in the announcements and it's really awesome, the model's been great. You have your own frontier model. Some people are saying, "Well, they have their own frontier model? Does it work with other things? What about model choice?" You guys have an architecture that allows for cross-cloud and multi-cloud or supercloud, which we called it, capabilities. So talk about this choice because you have Gemini. Great. That works for Workspace, CLI, all good, agents, but there's still that core interoperability heterogeneous piece. Talk about that because this is a nuanced point, but I think it's important to highlight and clarify. Are you guys locking folks into Gemini with the platform or not?
Karthik Narain
>> While we have always focused on creating the full stack capabilities so that we give customers the fastest and the most efficient ways to outcome, like the way the entire Google first-party surfaces are taking advantage of the full stack. So we are providing that to the external world, but we've always been open to other products. For example, almost all the frontier labs use, or majority of the frontier labs use Google Cloud as the platform in which they are training and inferencing. They're using our TPUs and GPUs from us sometimes. And they're also using the agentic data cloud platforms because they are dealing with so much data themselves, right? So we are making sure that we are not keeping the technology that is powering Gemini proprietary. We are giving that to everybody and saying, "If you have the talent to build the next generation models and products, we will give you all these capabilities that we have that's helping us build Gemini for you to build them." So that's one level of openness that we are doing. The second thing is across the stack, we are open. Look at the model serving. The new Gemini Enterprise agent platform is an improved and extended version of Vertex AI. Vertex AI is one of the most popular inferencing platform for Anthropic's cloud models, as well as several open source models, right? And that will continue.
John Furrier
>> So customers can do their own models with Vertex and Gemini-
Karthik Narain
>> Absolutely. So they can also bring a open source model and fine tune on our stack and they can use Gemma, which is built on the same architecture. It's not just as capable as the frontier models and they could do. So all of that choice using an open source model, they could make it their own. And even when you saw in the main screen when we said the models that are available yesterday, we talked about all the Gemini models and we also mentioned Cloud47 is available. We are putting that in the same plane to make sure that's there. Then if you go to the data layer, we have been the most open. Yesterday through the knowledge catalog, we are saying you can create business semantics from not just data stored in Google Cloud, but on AWS and Azure, Palantir, Salesforce, SAP, ServiceNow and Workday. And like I said, all these 350 portfolio companies of the Thoma Bravo and Vista are also going to be there. Last but not the least, Wiz that is now part of Google Cloud, their claim to fame and their success is that they're a cross cloud defense security platform.
John Furrier
>> I love what you guys are doing, amazing work. Now I got to ask you, you also run product and the go-to-market, which is very unique. You see what's going on in the product management side, you got the go-to-market with the full stack, how do you bridge that? And how do you optimize your time? Because you got the customer facing role and you got the product facing role. You're sitting on both sides of the table.
Karthik Narain
>> Yeah. Yeah.
John Furrier
>> What are you optimizing for?
Karthik Narain
>> I think there is absolutely a necessity in today's day and age that we build products that are almost ready to deliver... Again, the key word is almost, almost ready to deliver the outcome for the business. There is a changed expectation in enterprise buyers, C-suite, and that's why the business leaders are engaged, not just the IT leaders, even though the IT leaders role is even more important now as the engine for execution, but they are talking to us and saying, "How do I increase the share of wallet from a customer?"
So when they are talking about all of these things, we need to make sure that our products are designed to deliver those outcomes. For example, we are not talking to every customer and taking every request of theirs and adding it as feature, but we are seeing meta signals from them. When talking to customers, I'll give you an example. Our Gemini Enterprise for customer experience is used by customer service and call centers and customer contact centers. It was performing extremely well on chat interfaces, but we also provided voice capability. And in voice, the experience was dropping because the quality of a long context conversation with a thick context of a lot of information that needs to go, the quality was dropping and we recognized that converting speech to text and feeding the text to the model, getting text out of the model and converting that to speech was creating a quality degradation. This is an input from this use case or scenario that we are using for retailers and telcos and so on. Based on that, we developed a audio to audio model, and suddenly the quality of that is improved. So that is what I think-
John Furrier
>> So big feedback loops from the customer directly into product.
Karthik Narain
>> Absolutely.
John Furrier
>> Tightly coupled.
Karthik Narain
>> Absolutely. Absolutely. And the second thing that we are also doing is our sales team is transforming from selling SKUs, and products to putting together solutions. Customers are talking, we were with a large insurance C-suite and the insurance company is talking about, how do I change sciences where instead of the traditional way of using claims information and zip code data and prior pricing and saying, "This is what is the Monte Carlo simulation saying," they want a lot more multimodal information, whether that is satellite imagery of whether there is forestry here or there is flood path here, there's weather information, real time weather information, and determining, creating a digital twin and not just a probabilistic approach, changing each variable to see what will be... And it's actually benefiting customers with real time pricing and very individualistic pricing. You need to understand all of these things and put together the solution. Customers are not allowing companies to just give them SKUs and products and saying that all the onus is upon us. We're even taking this one step further with our forward deployed engineers to make sure we pick certain customers and we are going and extending our product because they are leading indicators of where the future is so that we bring back that extension into the core product.
Alison Kosik
>> Karthik, what's your biggest priority going forward?
Karthik Narain
>> I think there is three top priorities that I'm currently working on. Number one is deliver all the great announcements that we have made. Several of them were preview and we have a roadmap within the next month to two months for it to GA. That's priority number one because there's a lot of excitement and everybody wants these products yesterday.
John Furrier
>> There's demand.
Karthik Narain
>> Yeah, there's huge demand. So that's the thing that we are doing. The second aspect that we are doing is to constantly look at the improvements that we are making in our own Gemini model because the way we are, we are in a privileged spot where the products that we are creating, we are creating those products almost like scaffolding over the model. We believe the model to be the product. So we need to keenly understand the physics of the model. So as newer versions of our models come, we need to clearly understand what the new capabilities are. And based on the customer demand, we are feeding that to make sure the post-training phase is making improvements in those products. So these are the two product related aspects. And from a go-to-market, we want to make sure that we help customers accelerate their .
John Furrier
>> In the old world, they used to call that reverse compatibility. You're almost doing forward compatibility.
Karthik Narain
>> Totally.
John Furrier
>> You're looking at the model saying, "Hey, they're going to change." That's a static feature. It's always changing.
Karthik Narain
>> It's a feature, not a defect.
John Furrier
>> Yeah. Yeah. So you lean into it and say, "Okay, let's build around it so we're adjusting to the models."
Karthik Narain
>> Yeah. So the way we look at it is, a system is part model and part the software which acts as a harness. So as the software's capability improves on some dimension, the harness simplifies. In certain other dimension, when the capability regresses, the harness increases. So you could do this only if you have both a lab as well as-
John Furrier
>> And you got to scale. Karthik, we're going to have to bring you back and do an hour deep dive because you got a lot going on. Congratulations on the success and the product, and it hangs together. Everything's looking good. We'll see how it goes. We'll be tracking it. Thanks for coming on.
Karthik Narain
>> Thanks so much.
Alison Kosik
>> Thanks so much for your time.
Karthik Narain
>> Thanks a lot.
Alison Kosik
>> And you've been watching theCUBE. The leader in live technology coverage. We'll be right back.
Alison Kosik
John Furrier