Vidhya Srinivasan on Agentic Commerce and the AI Rewiring of Marketing

0:00Having deeply personalized experiences and having agents that are really helpful, assistive in getting tasks done will be something that we will make a lot of progress on. What we are fundamentally doing is reducing the commute cost from going from desire to actually getting what they want. And the tech can do that and therefore we will figure out a path because there's a real need for that.

0:27Engineering and marketing are often seen as polar opposites. On the engineering side, you have a world of logic, precision, and science. On the marketing side, a world of emotion, storytelling, and creativity. Today, those worlds are merging. Marketers are becoming technologists, and engineering has to get creative. Hi, I'm Josh Bania, VP of AI and Marketing Strategy at Google, and welcome to Frontier CMO. Today, I'm joined by a leader who, as much as anyone, is building the marketing industry. Vidya Srinivasan is the head of ads and commerce at Google. If you want to know how the entire advertising stack is being rewired by Gemini, or why universal commerce protocols will soon dictate how your brand shows up in an AI-first world, you cannot afford to miss this.

1:12Here is my conversation with Vidya, the architect of the modern marketplace.

1:20I think you're one of the most important people in the entire marketing industry globally. I don't know if everyone realizes that. So why don't you just explain for everyone what it is that you do at Google? So I'm responsible for ads and commerce. So responsible for the ads product, wherever it shows up, such as YouTube display and everything that we do in advertising. And on the commerce front, I'm responsible for shopping in search, as well as the Gemini app, Google Merchant Center and payments. So this includes Gpay and wallet. So that's a lot. So let me translate that into my head. Anytime anyone in the world runs an ad through Google or YouTube, search and beyond, you're responsible.

2:04Yes, pretty much. That's kind of an awesomely big responsibility. I mean, you have tens of thousands of engineers working for you. It's a billion upon billion upon billion dollar business. Can you just talk to the scale and how you manage a team that large with that much responsibility? It kind of blows my mind as a simple marketer. First of all, it's a phenomenal space to work in. I feel quite fortunate to work on the problems that I get to work on. It's actually quite fun. So as much as it's a big responsibility, and I don't take that lightly at all. At the same time, I really, really love what I do.

2:40and I'm inspired by the impact that we have and I'm inspired by the people I work with day in and day out. It is a big responsibility and it is one that has far-reaching consequences in the world in terms of how it affects businesses and how it affects consumers. So Vidya, you publish a letter every year to the advertising community, to marketers, to everyone about the future of Google ads, YouTube ads and commerce. This year seems to be a lot about agentic commerce and the UCP protocol. What do we need to know? The best way to think about agent e-commerce is for decades, a person could either shop fast or shop smart.

3:20And with agent e-commerce, you essentially don't have to choose because the straight off that one has to make between speed and certainty is being reduced with AI. So stepping back, the vision really is to make commercial experiences across the board assistive, personal, and fluid. and businesses have an entirely new way to engage with consumers where they get to bring the full breadth and depth of their offerings but do it in a way where people don't have friction in the process. Now UCP, Universal Commerce Protocol that we announced in January this year is one of the foundational pieces that will help this thing actually become a reality and you can think of this as a kind of a common language that retailers and merchants can use to connect to and talk with AI agents that want to participate and facilitate commerce.

4:14Obviously, Google search has been going through its own unbelievable transformations, more innovation last year than maybe in the few years before that. Right now, people are writing much longer queries in AI mode and AI overviews, which is giving us a lot more context about what they're interested in, what their intent is, right? It really is an expansionary moment where people are asking us longer questions, deeper questions and we can now bring to bear everything that we know about the world, but also everything that we know from a commercial point of view from our shopping graph, all in one place.

4:46From an ads perspective, it gives us unique opportunities because we are using this moment as an opportunity to really reimagine and reinvent what ads are for search. We can't talk about ads without really talking about YouTube. Tell me a little bit more about what's happening with YouTube and YouTube ads experience. YouTube has been on a phenomenal journey where we really see how the YouTube creators are sort of the center of creating culture and being tastemakers. And obviously, this is something that brands care a lot about because finally, it's about influencing people and finding the audiences that really resonate with the brands.

5:25We've been very focused on helping brands and helping advertisers in general find ways to connect with the right communities through the right creators. We announced the Creator Partnership Hub some time ago. And you can think of this as just the command center where you can go and do all things creator oriented in finding the right creators and understanding audiences and having the measurement tools and understanding effectiveness. and we continue to invest in new features and products within this space to help advertisers and to help creators. So just different ways to matchmake across that marketplace.

6:01We have Gemini. It's the foundational model of Google, leads all the leaderboards. How's that translating specifically into ads? You mentioned PMAX and AI Max for search. Talk a little bit more about how Gemini is really up-leveling these products themselves for advertisers. We use Gemini everywhere. I would be hard pressed to tell you where we don't use it. We use it as I was just talking about. We use it for all the intent matching when we are for the ads quality stuff. We use it for, of course, generating the creatives itself. But then we also use it to drive all our performance campaigns. And as these models improve a couple of times a year, every time we get a new version and we bring that to bear within the ad systems, we get a lot of gains just from having better models at play.

6:53So always part of your letter, core to Google's DNA is the notion of trust. We have to build a system which works for everyone, bold and responsible ways in Sundar's language, but just trust is at the heart of what we do with our ads business as well. Any sort of speed or innovation, nothing is useful without safety and trust as a foundation. We have a very strong legacy of building products that are highly trusted by consumers for the last 25 years. And as we look at all of these innovations that I just spoke about, even in the AI era, it is all built on this foundation of continuing to preserve trust with our consumers.

7:34And a lot goes into that. There are many, many things we have to put in place for this to work well in a scaled fashion. And that continues to remain a top priority for us, even as we embrace AI wholeheartedly. What is the thing that CMOs just don't quite understand that you can explain to them now or you wish they did understand? I think what CMOs may want to embrace more is the fact that so much of this magic is happening from a foundation of curiosity and optimism. and a framework where you're able to test fast, learn fast and iterate fast. So shut things down that are not working, learn something and then focus on things that are working.

8:12And sometimes there is a desire to know things for sure, but predicting the future accurately at the pace at which things are changing is going to be hard. So I think at this point, sort of embracing that as the opportunity and honestly having fun. I think that's a big part of it. But I just, it is fun to also have unknowns. It's fun to sort of explore, co-create sometimes and to figure out what's possible. You're an engineer at heart, right? You're a builder. We'll talk about that in just a second. But talk a little bit about your systems. Like, do you have any frameworks or principles or guidance of how to just keep this all straight in your head and how you run such a large organization?

8:52I think of myself as a values first leader and values help when you have trade-offs to make. and when you are in these moments where you have to pick left or right and then what are the guiding principles to make those calls, I then work with leaders and actually pick leaders who are aligned on that front. And that becomes a common ground for me and my team as we navigate some very, very tricky issues in these areas. When you say values as a marketer, I tend to think of values in the sense of almost moral judgments and we're going to stand up for this and that. Is that what you mean as well within an engineering context or is there other ways to interpret word values?

9:30I think it comes up in many different contexts. So for example, one of the core values we have is always user first, do right by the user. Yes. When you come to those points where you're making a product decision, could be an engineering decision, it could be something else in the UX, any which way, there is a very strong north star to what you're going to optimize for. I believe in fairness and it matters a great deal to me that we do our best in ensuring fairness when you lead teams. And that manifests itself in many ways in being very conscious about processes that don't have bias and things of that nature.

10:08Another thing I believe in, and I still think of it as a value, is the ability to have open debate. Ability to do conflict in a respectable way. I feel that's a core value because I feel if I create an environment within my team where all my directs minimally can openly discuss any conflict that arises and they can quickly get a resolution. It unblocks a whole lot of decisions further below. So the speed of decision making is one thing that you have to really, really, really focus on. Having the space to resolve conflict is a big part of that. In fact, one of the things I say to my teams is at core, the way I lead is to deliver customer value at speed and speed has so many aspects resolving conflict is one thing accountability and ownership is another thing when we're creating teams and we're creating organizations that work with each other having clear lanes for people is another thing that I focus on a lot and all of that goes into the velocity of the products that we're able to build so there's this really interesting discussion happening on the marketing side of the business right now and really it's around do marketers need to become technologists and can you be a successful CMO if you can't code what's the minimum viable amount of engineering speak a CMO needs to have today and going forward that they might not have needed five years ago I think the bar for how deeply technical and how much you have to understand as a CMO has gone up significantly and that's largely to do with the fact that you know traditionally you would say you've grown up a CMO's traditionally grown up as a marketer and during that process this person has gained a lot of intuition about what feels right what feels wrong and the team comes and presents a set of metrics and they can very quickly know okay are we on the right direction or not is this the right channel is this the right allocation the problem now is all of those things are evolving and changing.

12:06And so I would say for a CMO, you kind of have to have two piles. One is a pile of things that you know really well. And that ends up being things like, what are your goals? Like, what are the things that you're optimizing for? Who's, what, how should the brand feel like to a user? What should be the values that you're trying to project? So those things, I think the CMO has to know super well. And then the how you go from, you know, wanting the metric to actually getting it, there's a lot of unknown now because there are lots of new platforms and many new tools that they can use to achieve these goals and so then the question becomes how do you tackle that set I don't think a CMO needs to code I don't think the CMO needs to know deeply the algorithms of every LLM out there or know exactly how it works internally at the same time I think the CMO does need to know at a high level conceptually what what is the different technological pieces available.

13:02What can you do with creatives? What can you do with, you know, coming up with briefs? What can you do with product development? How do you brainstorm? How do you, at a high level, you should understand how can things speed up coding? And the CMO needs to know how these things can fit with each other. I would ask you to be a little vulnerable here. What is it something that you don't know the answer to yet, but you're working on and trying to solve for Google's customers? There are many, I would say. like even if you take an area like measurement, there are many things that we know and we model it very accurately.

13:36But what I constantly hear from CMOs is, hey, we want even better measurement. We need to be able to compare two or three things that we currently don't have an ability to compare well. And then you have different ways to model this and do things. The thing that we will not compromise on is we're very particular about only giving statistically significant results and really standing behind what we give. But there are many things that we would like to produce results for that we hold back because we're not sure how far we can go on that. And those are areas we keep trying. We are coming up with different attribution models that will help map more closely to what marketers expect or the way they think about it.

14:19So on the marketing side, we want certainty. We want to know that our incrementality and attribution worked. but we want to know it across everything everywhere all at once but it's also very challenging when your cfo says deliver this impact and you can't yet prove it to a absolute certainty that's one of the things we're just hoping to try and get better at and better and better at i guess one of the things that's interesting i'll stick with measurement there is a meme a thought that the classics of measurement you know awareness consideration intent, media mix models, that the way of doing measurement has been relatively slow, relatively stable, and it kind of works in a very finite way within maybe packaged goods.

15:01But the speed of business, the speed of decisions, the breakdown of the classical funnel, people see something, a YouTube short here, and then they see a YouTube video there, and they see a search ad here and everything else beyond that, instant purchase. Is there any hope for actually cracking the speeding up of measurement in your mind? Absolutely. I think we are in a position where the way the models are able to predict and keep up with the changing user behaviors are pretty phenomenal. This is why automated campaigns became so successful. I mean, this was a hypothesis. And then what we actually saw is it was very much verified by usage and what advertisers were actually seeing day to day.

15:42Increasing the world of things that we know for sure, the ground truth, the first party data around conversions and other things are very important. So I think this problem, we can do a lot better if advertisers sort of hold hands and put a lot of energy into collecting, having the consent for appropriately leveraging that data and having all the ways of feeding that to the campaigns because we can go quite far with that. Come to the table with your first-party data. And a lot of businesses still do not have their hands around all the data they have. But is that one of the key unlocks that marketers should hear?

16:20Absolutely. As AI gets better, it's not going to work for you. The models might be better, but it's not going to work for you unless it has high-quality data to work on. We've talked about first-party data. We're talking about AI. we're talking about quality data sets. What about the role of creativity, the magic, the art? And I guess as you've been in this role, as you run these teams of statistically, mathematically minded engineers, what gets lost in translation from the marketing side to the engineering side around that part of it? So sometimes there is this false notion that the tech can do more than it really can.

17:04so much of the fundamentals is still in the hands of people certainly for creativity taste and ingenuity it's entirely comes from from people from good marketers it's good to hear that's the foundations of everything else that we do in terms of what is good and bad if you think about how ai predicts what comes up with the right creative which is which is good for that context well how do you define good well good has to be defined by a person so that part is still very much very much needed and in fact I would say it's needed even more because there's going to be a set of things that can that you can get as tailwinds once you know that for sure like if you know what good looks like and what great looks like for your brand there are a bunch of things we can do around that to just make it to modify just a little bit here and there to customize that look AI comes up with options people still have to pick the direction so judgment is a very very critical thing to hone especially at this time so i want to pivot just a little engineering writing code is a creative pursuit heavily math influenced i get that but when did you realize where were you what were you doing when you thought i'm good at this i love doing this i'm going to make my career to become a builder of code and systems and architecture of these sorts of solutions this happened actually quite early for me uh i remember it it's one of those moments you remember from your childhood when i wrote my very first program i think i was in 10th grade i wrote something from from a book i copied i typed it out uh and then when i ran it it actually worked it was a pascal program okay we had a very very arcane system to run on uh but still it was um incredible to sort of see that it actually worked and then you could create different functions and sort of like lego blocks you can build different things connect them and then more things happened that was I think two things one I was just thrilled by the fact that it would happen because it was the first experience but second thing is it felt natural to me it kind of felt like this is how I think and And I could translate it to code in a more natural fashion.

19:22And so that's how that started. You actually spent a lot of time working on large databases and building out databases. And as someone who's non-technical, it's all like Egyptian hieroglyphics to me. But when you joined Google six years ago, you had never really worked in ads or you were fairly new to it. Just talk about the learning curve of going from one type of engineering building large databases to a different type of engineering building for Google's ad products and what that experience was like. To be candid, I did not realize that the ad space was as complex as it is. So I had the benefit of not knowing that, I think, when I took it on.

20:04at the same time though i think when i in these sorts of transitions i think the key thing is to look for adjacencies like the things that you know that you can bring forward and then the things that you will learn in terms of adjacencies at its very heart ads is a big data problem in fact our ad systems were some of the largest data systems even within google we leverage machine learning earlier than most other other domains within the company and so there was a lot about the the guts of the ad systems if you will that felt very natural to me those things I could quickly understand I think the parts of advertising that is hard that took time to to sort of fully appreciate is just how expansive it is as a business in how it interfaces with the ecosystem that there is a massive interconnected network across the board in every website, publishers, advertisers, now creators.

21:04And of course, there are lots of third parties involved in various, various aspects. I think for me, understanding how these things are correlated and how they can impact one another, at least having a strong hypothesis, is both interesting and important. AI is transforming our business. Obviously on the marketing side, where everything's being rewired and rethought because of it. It's happening on the engineering side. I'd love to just get your thoughts about how are you thinking about AI to transform your ways of working inside of the engineering organization? And then I want to try and flip it and talk about how we might do that in marketing.

21:38Obviously, this has been a journey. And I think we all need a new playbook when it comes to this. It's an evolving thing because so much is happening every few months. I would say sort of maybe five things I can touch on And as principles, I've sort of found to be effective in leading organizations through a massive change, like what AI is for us. The first one is you have to lead with optimism, not fear. Focus on the opportunity and honestly, the fun parts of embracing this. I'm not saying that there will be things that we find that are not very positive. But I think focusing on that upfront leads to no, it means it's just not very useful.

22:21The second thing I would say is clearly have people, the human to be in charge of all the key decisions. Having key decisions be with named individuals as you're trying to do this transformation, I think goes a long way. Could be in terms of the final checkoff for code reviews. Every organization has their own set of processes. I think start with people being in charge and having the AI do a lot of assistive work so that you get time back. right that has helped um i would say the third thing is really celebrating all aspects of the experience so we tend to celebrate the successful use cases right when something nice happens oh great we got this it's great but um in this world a lot of a lot of things you try is not going to work and so making a habit of also celebrating the things that didn't work because the learnings are almost more important than the things that worked at this phase i think that that goes a long way because everybody wants to be on things that mean something.

23:19Another aspect is as much as there's a lot of hype around it, I think it's important to make it practical and real. So sort of stay away from the hype. And the way I found it effective is across my organization, we try to come up with areas and specific projects that has specific outcome goals and it's time-bound and specific set of people and have them go work on something, some new piece of new way of doing it and then come back with to discuss what happened. There's so much talk about what's possible in the fullness of time or what's possible maybe in the context of a new startup or the context of a brand new code base.

23:59Well, most of our code base is not that. The fifth one is actually for this sort of thing that we're trying to do. Of course, there's some amount of, you know, tops around messaging about, hey, I think this is important. and we should try to do this, we need the tools, etc. But I think behavior leads to culture more than any other way of leading culture. So I think the more the leaders and people who want to see this change happen, embrace it themselves and talk about it, the more it has a snowball effect across the rest of the organization. What have you told your team about some examples of where you've been using AI to do things in your own life?

24:38I build little applications with my daughters, for example. while, you know, they have lots of activities and they wanted a calendaring app. They wanted some extra features associated with that for notification and things of that nature. So very, very simple, but it was just a way to do something with them, but also understand how easy it is to do with the tools that are now available. Okay, that sounds absolutely awesome because we all need better notifications. So how long before we can productize this and push it out to everyone in Google Calendar? I will let the workspace team know that there's a latent need for this.

25:14I'm just saying your family, they're the customer. The customer is never wrong. That's right. 2025 was a banner year for Google's technology under your watch. I think 2026 is going to be even bigger and better. What are you really excited about in this AI moment and transformation of all the products that you help lead and build? In 2026, you know, it's cliche for me to say this, but I think the holy grail of marketing has always been the right ad at the right moment for the right person. And I think we are closer to achieving that for real than we've ever been in the past. I think we're going to make tons of progress on that front as businesses fully embrace what tech can do for them.

25:55and they get incorporated into the platforms and this whole flywheel fully takes off. From a consumer standpoint, I would say having deeply personalized experiences and having agents that are really helpful, assistive in getting tasks done will be something that we will make a lot of progress on. How far we go depends on lots of the ecosystem coming together. But certainly that is another area that is just ripe to getting right. because what we are fundamentally doing is reducing the commute cost from going from desire to actually getting what they want. And the tech can do that. And therefore we will figure out a path because there's a real need for that.

26:38So across Google, across YouTube, across commerce, across all the ad systems, we're moving from a era of, wow, that'll be exciting to, here's how you actually do it. From wow to how, thanks to all of Google's and YouTube's ads. That's the plan. I would ask you some signals or noise. Give me your quick assessment. Okay. Okay. Open protocols will matter more than platforms. Signal or noise? Signal. We got the signal loud and clear when we launched the Universal Commerce Protocol. Agent to agent commerce will become mainstream in 2026. I think more noise. I think lots of the building blocks are being laid out.

27:16I think we probably have some time for mainstream. Star Wars is way better than Star Trek. signal or noise. Oh, signal. No doubt on that one. Thanks for coming, Vidya. It's been a great conversation. Thank you for having me. Before we wrap, I want to leave you with three takeaways to carry into your week. Here are your field notes from the frontier. First, Vidya's message is clear. The holy grail of marketing, the right ad for the right person at the exact right moment is no longer a theoretical dream. It's within technical reach. CMOs don't need to write code, but they do need to be conversant in the language of technology and what these new tools can do.

27:52Second, we are moving beyond consumers having to choose between shopping fast or shopping smart. Agentic Commerce is merging these two needs and personalizing experiences. AI agents can facilitate the entire journey from discovery to payment with zero friction. Third, Vidya charts a course for CMOs through a five-point leadership blueprint. Lead with optimism, not fear. Maintain human accountability for key decisions. Celebrate failures as much as successes. Don't get distracted by the hype. Stay focused on grounded business problems. Finally, create the culture change you want to see by using the technology yourself.

28:33Vidya models this by using AI alongside her teams and even building simple apps with her daughters, proving that true inspiration comes from seeing exactly how the technology works firsthand. This is the work at the frontier. Thank you for joining us. For research, case studies and tools to help you lead through this era, visit thinkwithgoogle.com where we track the signals, cut through the noise and share what leaders are actually testing in real time. We'll be back next episode with another leader operating at the frontier. Until then, stay sharp and stay out front.