State Farm’s AI Reckoning

[00:00:00] This episode is sponsored by HPE. Today's networks must evolve from simply providing connectivity to harnessing AI to deliver exceptional connected experiences. At the break, HPE's Rami Rahim shares how self-driving networks can make that happen. Welcome to WSJ Tech News Briefing. It's Friday, July 10th. I'm Imani Moiz for The Wall Street Journal.

[00:00:24] Tech giants have long highlighted efforts to reduce their environmental footprint, but many companies are using far more water than they report. We're breaking down why you can't always take a sustainability report at face value, and why tracking how much water data centers actually consume isn't as straightforward as it may seem. Then, is Jake from State Farm going to be replaced with a chatbot?

[00:00:48] We'll explain how the century-old insurance giant is reinventing itself for the AI era, and why the changes are sparking backlash. But first, AI is driving one of the biggest infrastructure buildouts in history, estimated to cost $1 trillion for this year and last. As communities weigh whether to welcome new facilities, measuring their environmental impact is becoming increasingly complicated, particularly when it comes to water.

[00:01:17] Water consumption is projected to grow rapidly in the coming years, and there's no law that requires companies to report the full scope of their water use. So annual sustainability reports often don't paint a complete picture. WSJ columnist Christopher Mims joins me now to explain why the math doesn't always add up, and what you should know if you live in a water-stressed area.

[00:01:39] So large tech companies report how much water they use in their annual sustainability reports, but you found that those figures don't tell the whole story. Why is that? Large tech companies only report what's called direct water consumption. What they're not reporting is that there's another type of water their data centers are using, and this can be 10 times as much water as they're directly using in the data centers.

[00:02:04] And that is the water required to generate the power for these data centers. And it's a big deal because so many of these data centers, especially the new ones, are powered by fossil fuels, which require quite a lot of water. So why is indirect water usage getting more attention now? Companies have been kind of trying to, I wouldn't say hide it, but they don't report it, with the exception of Meta.

[00:02:30] One reason indirect water use is starting to get more attention now is that many data centers are being placed in areas that are already water-stressed. And frequently, the power being used by these data centers is being generated by power plants that aren't too far away or may also be in water-stressed areas. And some companies say they're offsetting their water consumption with other sustainability initiatives. How does that work?

[00:02:58] So a lot of companies are claiming that through so-called water restoration projects, which could be anything, maybe they restore a wetland in order to retain more water, or they're paying to stop leaks in a municipal water system, that they are offsetting the water that they use directly in a given community.

[00:03:22] And so they're claiming that by 2030, Google talks about this, Amazon talks about this, they're going to put back a gallon of water for every gallon that they directly use in their data centers. But again, this doesn't account for indirect water use at the power plants powering these data centers. So how does the location of a data center affect water consumption?

[00:03:47] When a data center is in a water-stressed area, and when it is drawing power from a water-stressed area, it can draw down local supplies of water because water is regional, right? I mean, water travels, but ultimately if you are drawing water from the Colorado River and the desert southwest, there's only so much water available. Are there efforts to standardize water use or consumption reporting?

[00:04:17] It seems like there's a lot of different ways that this can be measured. There are a lot of different ways to measure, but what we don't have a handle on yet is how do you look at the sort of all-in water consumption of a given data center in a given location? The challenging part is every single one of these is different. The location it's in is different. The climate it's in is different. The mix of power going into it is different.

[00:04:45] So it's a big job. It would have to be calculated individually for every single data center. And of course, it depends on the weather. It depends on the time of day. So Christopher, what's the big takeaway here? If you live in a water-stressed area and somebody proposes building a data center in that area and they say, hey, don't worry, we're going to limit the amount of water it uses, the next question should be great. Where does the power for that data center come from?

[00:05:14] Because if it means that, let's say, a coal-fired power plant that will be retired is going to stay online. If it means a bunch of new natural gas generation, which is often the case, the actual all-in water use of that data center could be 10 or 20 times greater than what is reported. That was WSJ columnist Christopher Mims. A quick reminder that we want to hear from you. Have you seen an AI-generated post that you thought was real?

[00:05:43] Shoot us an email to tnb at wsj.com or leave us a voicemail at 212-416-2236. That's 212-416-2236. Or if you're a listener on Spotify, leave us a comment. Coming up, State Farm is stepping into the AI era, but it's having a hard time selling the strategy to its agents and customers. That's after the break.

[00:06:16] Here's Rami Rahim, HPE Executive Vice President, President and General Manager of Networking. A self-driving network is a network that essentially configures itself. And what used to take months, maybe up to a year to deploy, can now take weeks, if not days. A network that optimizes itself. And very importantly, a network that heals itself.

[00:06:43] Thousands of State Farm agents gathered in Las Vegas last month expecting a celebration. Then, right at the end of the convention, they got unexpected news. That an AI overhaul was about to fundamentally change their jobs. And their benefits. State Farm CEO announced a new contract. It asked health benefits and outlined a new compensation deal, sales targets, and the addition of AI to the sales process. Backlash from both agents and customers was immediate.

[00:07:12] And that's raising bigger questions about how companies can embrace AI without losing the human touch. WSJ's insurance reporter, Gene Eaglesham, is here to unpack the fallout. What does this AI overhaul look like in practice? Why is it such a big shift in State Farm's business model? State Farm has for decades had this pretty old-fashioned business model where you have a sales office in pretty much every town, sometimes several of them.

[00:07:40] And people go there, they buy insurance, they get to know their local agent. What the company is now saying in a competitive pressure is it needs to respond to the fact that customers want to do more themselves. They want to be able to buy things online. They want faster, more efficient service. So they're using AI to transform really how you can buy insurance to give the agents new tools to try and modernize the whole process.

[00:08:09] How did agents react to that plan? Agents have been stunned by this. We've heard from scores of agents and words we hear again and again are betrayal, broken trust. They feel blindsided by it. So agents, they're in a curious position with State Farm. They only sell its products, but they're independent contractors. But they thought that the benefits they had, they thought those were effectively guaranteed.

[00:08:35] And suddenly State Farm has said, no, that's all ending. We're forcing you onto new terms. And it's a kind of take it or leave it deal. And there will be a payoff for those who choose to quit. But agents were saying the amount of the payoff is just not that high. They feel it was poorly communicated. What we've seen from internal State Farm documents now is State Farm says it's in listening mode. It is talking to agents. It's reconsidering some elements of the package.

[00:09:04] But it says the general direction it's set on. It's going to continue with this. It says the changes that are being made are necessary to effectively future-proof the company. A lot of the agents aren't convinced that State Farm's AI is ready for prime time. What kinds of problems are they running into? One of them said it's a great insurance company. It's a horrible tech company. And a lot of big insurers are like this. They have old systems.

[00:09:31] They're layering new tech on top of software that maybe dates back decades. And the new AI tools, what they've seen so far, there are some glitches with it. So, for example, one of State Farm's biggest new products is something called Household Story that gives a snapshot of every customer's household, who the people are, what their financial position is, what their needs are. And agents who've tried it out say, in some cases, it was inaccurate information

[00:09:59] from the day you became a policyholder, for example. And we've also heard at the moment State Farm is an AI assistant for agents. And one of them asked who won the last Super Bowl. And the answer they got was from 2024. Broadly from agents heard is that people want to be able to use the advantages of AI to do things faster, more efficiently, more effectively. They want the new tools. They're keen to use them. It's just this question mark.

[00:10:26] And this issue for all these big companies, really, is how do you make sure you roll it out in a way that's very effective and make sure these tools actually work? And one of the headaches of State Farm right now is they need the agents to help with that. But at the same time, a lot of the agents are up in arms about the changes to their pay. You heard from over 1,000 readers after your story came out. Many of them were State Farm customers. What do they think about the changes?

[00:10:53] What came clear from the responses is that the technology the insurers have right now, particularly chatbots, are driving people crazy. We heard again and again people hate being stuck on these automated answering systems. They don't like the tech that the insurers have at the moment on their website broadly. They want to deal with people. They want to be able to reach a person. Do you think the response your article receives says something broader about where we are as a society when it comes to AI adoption?

[00:11:22] A lot of people are really worried about how AI is upending life, how it's automating so many things, how it's impacting relationships. So most of the responses said they accept this was inevitable. They know it's coming, but it doesn't mean they like it. And I think people really want to keep human relationships where they can. And that desire is not necessarily being reflected in what executives at a lot of big companies are doing.

[00:11:51] Okay, I have one last very silly question for you. I always like this. What about Jake from State Farm? Is he going to be okay through all this? Jake from State Farm, I think, will still be there. He was at the convention. The actor playing Jake was there. A lot of happy selfies with him with agents. So I think State Farm is saying they're going to keep the human element in this. They recognize it's important. That is still key to their sales model. So they're going to try and keep that. Jake is still going to be there. Whether we get an AI Jake, we'll wait and see.

[00:12:21] That was WSJ reporter Gene Eaglesham. And that's it for Tech News Briefing. If you're a listener on Spotify, be sure to leave us a comment. Today's show was produced by Julie Chang. I'm your host, Imani Moiz. Additional support this week from Danny Lewis. Jessica Fenton and Michael LaValle wrote our theme music. Our supervising producer is Katie Ferguson. Our development producer is Aisha Al-Muslim. And Chris Sinsley is a deputy editor of audio for The Wall Street Journal. We'll be back later this morning with TNB Tech Minute.

[00:12:51] Thanks for listening. Here again is HPE's Rami Rahim to explain why it's critical for modern enterprises to use AI-powered self-driving networks. Gone are the days where humans can keep up with the complexity and overcoming the cyber attacks that are happening using elbow grease.

[00:13:16] You must leverage agents and artificial intelligence in order to deliver a seamless connectivity experience to everyone and everything. It's not just people that are connected to the network today. It's AI agents. It's billions of things, in fact. You can't do this manually. It must be done using artificial intelligence.

[00:13:38] And everything from the deployment of the network or speed matters to the identification of issues that inevitably happen to the remediation of those issues needs to be done without human intervention. Find out how a self-driving network can help you deliver exceptional user experiences at HPE.com.