Welcome. I'm Ena Freed, chief technology correspondent at Axios, and I'm so glad that you're all here, whether it's in our room here at Axios House in Davos or following along on the web. Today we're going to have a really dynamic set of conversations around where all this AI stuff is going. We're going to look at the building blocks, whether that's dynamic grids, clouds, all that sort of thing. What it takes to power all the places we know we want to go. And with that, I'd like to welcome our first guest, Thomas Kurian, the CEO of Google Cloud. Thomas, so it's been a pretty good few months for you. You know, Google's cloud, we're hearing a lot more about it. Talk about obviously, you know, I don't think you were as out of the race a couple months ago as people said. I don't think the race is over, but what's changed in the market dynamics? What's fueling the momentum that Google's cloud seems to have?
I think we've been working on this for a long time, which is to do AI well. We've built a vertically optimized stack of technology from our own chips, our own models, our own tools and applications on top of it. And then we've integrated it into our scaled consumer services, so search, ads, YouTube. We've also seen huge pickup for our Gemini app, and then we've also integrated it into our enterprise cloud services. We have close to 130,000 companies using it, and all that helped improve both the quality of the model as well as the foundational breakthroughs we were making. And so for us, it's been a three-year journey, and we're very confident not just on where we are now but our path forward as well.
And on that, well, we'll get to the future in one sec, but in today, you're meeting with customers, your competitors are meeting with customers. If I'm a CIO and I'm deciding between your cloud, AWS, or Microsoft Azure, what's one reason that you would point to that people should go to Google Cloud that maybe wasn't as true a year ago?
You know, at the core, we have our own chips and models. Just like in personal computing, to do it well, you either had to own, as Intel owned the processor, or somebody owned the operating system. If you don't own chips and models, you're just a distributor. And so we're the only one who have both.
And talk about where we're going because obviously AI is moving fast. I still feel like the technology underlying all this is changing faster than anything I've seen. What's something businesses will be doing with AI by the end of the year that they're probably maybe not even on their radar right now, but certainly that they're not doing today?
Broad brush, if you look at where people are using our AI tools within companies, they're doing four things with them. One, many companies are taking products and building digital versions of those products. So for example, if you're a manufacturing company, adding a control system with a human interface that can talk to your manufacturing systems using large language models. If you're a consumer services company, for example, creating digital tools around the model itself. So we've seen many people build agents for education, for health care, a variety of things and offering that. So one is really building a digital product. Second is transforming the interface between you and the customer. So for example, the way that shopping is working online is changing. We have huge interest on agentic commerce. The way that companies serve their customers online, the way the government serve citizens online, they're all changing using these tools because you can have them help people find information and answer questions in a much richer way than before. Third is transforming the core of the company. If you're a securities firm, improving settlement. If you're a bank, speeding up issuance of credit cards and loans. If you're an insurance company, changing the way you do claims processing. If you're a health care institution, changing how you're planning your surgical procedures and how you manage your hospital beds. And then lastly, in the IT organization, we've seen huge changes in software engineering, in cyber security professionals using AI to identify threats and start defending against them, and in the way that people are analyzing information.
So I think a lot of people come to Davos every year, and certainly since the arrival of ChatGPT, it's a lot of everyone wants to know what should I be doing, what shouldn't I be doing. Most of the companies, they did pilots two years ago, they tried to move things into production last year. Some are succeeding, some are getting stuck. So I want to talk about what you're seeing. So for customers that are succeeding, that are getting positive returns in AI, what are the one or two things that they're doing that are allowing them to succeed? And then we'll get to the folks that aren't having as much success.
Broad brush, the ones that are doing it well, one, they're organizing their data because if your data is not well organized, complaining that the model is not finding the right information doesn't make a lot of sense. Second, they ensure that they're not just using the model but they're actually transforming the core process itself rather than draping a model on top of an existing system. They're changing. Take a simple example. If you think of customer service, the leaders have changed where instead of having a human in front with a model behind it, having the model in front with a human behind it, and that drastically improves how well the model starts understanding the needs of the customer to help resolve the problem. So I would say the three things: organize your data, ensure you have a strong foundation so things like your security protections, cyber defenses, etc. around that layer is built in, and third, rethink the process where you're using and depending on the model to do more things as opposed to just putting the model as a sidecar on top of a human process.
So obviously, I mean, on the flip side, obviously if companies aren't ready with their data, that's one reason they stall. For other companies that get stuck, that don't see the ROI they were hoping for, besides data readiness, what's one thing that companies that aren't having the success they want tend to do wrong?
It's, I think, when AI is driven as a pure technology project, often it fails because the business line doesn't understand and doesn't agree on the metric that they're going for. So if you look at just customer service, just using that as a single process, some companies that we work with see it as improving retention of existing customers, or if you don't do it well, they're frustrated and they will churn. So for example, insurance claims processing, speeding up claims resolution is the best way to retain customers. Other people who have a broad product portfolio are really keen on using customer service to drive conversion and upsell on product. Now, if you're a single product company, that's not a key metric. But the way that you do upsell using a model is very different than the way you do resolution. And the third, still others say, well, you know, I want to use it as a way to optimize my cost to serve. So for example, if you're in banking, one of the challenges people have is serving the long tail of consumers with financial advice because the cost of having a human advisor doing it, it works fine at the top tier with the private banking, but at the long tail, you can't do it. But the tools are available to serve the long tail. Now you have to decide up front, are you focused as a customer on churn reduction, new...
You have to figure out what is it you're using AI for?
