OpenAI Announces New AI Tools at DevDay, Google Is Paying Digital Publishers for AI Overviews

0:13Welcome, everyone, to The Information's TI TV. My name is Akash Pasricha. It is Wednesday, September 30th. We are back in our New York studio. We are continuing to watch reactions today to President Trump's meeting with AI leaders at the White House yesterday. The executives signed what they call a morally binding deal around artificial intelligence, or what the president is now calling superintelligence. There were a lot of funny videos and tweets about the press conference, and it'll be interesting to see if the companies actually change their strategy at all from here. Today on the show, we are recapping OpenAI's Dev Day, which also happened yesterday.

0:55We're also going to discuss our exclusive reporting that Google is now paying certain publishers for content that helps generate its AI-powered results. We're then going to bring on the CEO of Elise AI, who just raised funding at a$4 billion valuation. And we're going to close out the show with a conversation with Marissa Mayer, the former CEO of Yahoo, who is launching Dazzle, which is using AI to analyze photos on your phone. It's going to be a great show, so let's get right on into it. OpenAI hosted its Dev Day yesterday, releasing a whole host of new products, including Dots, a new AI agent to take on Muse.

1:33Space, a new collaboration platform with Docs and presentations. It also released a newer, cheaper model. Ray Wong, the CEO of Constellation Research, was there on the ground. I want to bring him on to share more about what he saw. Ray, welcome back to the show. It's great to have you here. Always great, Akash. How are you doing? I'm doing well. So what was the reaction on the ground to Dots after they released it yesterday? It's not like too late, too little. well, we need more from you because they know the possibilities of what OpenAI can deliver. But there was a relief like, hey, we can do this in OpenAI.

2:10We don't have to go to meta. So I think that was the big thing. But dots, for those who don't know, these are your customized, persistent agents. They're specialist dots. These are things like, and unfortunately, I'm going to say it this way, they're meta muse for OpenAI, which is kind of what people were saying on the ground. So that's kind of a sad reaction. I think when people refer to your new product, in the context of another product. Tell me, what about pricing? And maybe I missed this detail. Do they have a pricing yet on Dots or is it included in the ProPlan? I don't know if we had pricing on Dots yet, but I think it's in your ProPlan and your business premium plans that are there.

2:50They also announced a bunch of plans. You have the new top tier$500 per month plan, which I'm probably gonna bluff through. There's the Pro 200 subscriptions are finally back. And of course, they're having better API times. It's 45 % lower, which is what people were cheering to. Right. And I guess when you say too little, too late, I mean, it sounds like people were underwhelmed even by what they saw. Like there wasn't a lot of confidence that it's going to be as good as what Muse or Instinct can do. Well, no, I think the concept was really what they were saying. I think they believe OpenAI can deliver better than Meta.

3:27And I think that's really was like, guys, you should announce this earlier. and I think people were basically saying, look, I've got new models, I've got new capabilities. I mean, Dots is really one of the things that was important. But it's the specialist Dots that I think people are most excited about, the ability to get to specific tasks when they want to. But on the new model front, right, GPT-6.1 Sol, along with a whole bunch of previews, I think that was exciting for a lot of folks who were trying to see what was next and what they had to prepare for. And of course, a lot of the API public betas, as we get geeky on APIs, decision APIs, And of course, as you were saying, spaces and marketplaces.

4:03Right. What about the, so let's go to the$500 a month plan. Do you think people are going to spend that much? Is this going to really be an enterprise product? How much demand do you think there's going to be? Well, that's the thing, right? What is an enterprise product if you are open AI? And the question is, it's by consumption. And I think for a lot of folks, it may be the$500 per month plan. I mean, this is the top tier. And so I would say that it's pretty easy to blow through$500 per month if you're building a lot. If you're creative, you're like a dev, you're coding like crazy, or you're actually having the agents code like crazy for you and you're managing multiple products at the same time.

4:45So I can see that happening already. I've got several friends that basically have like four or five agents talking to each other all the time and coding with each other all the time. And that's really what's going to drive that. Right. You know, I have to tell you, though, although they were not first to market with Dots, you know, I can't help but think that it doesn't really matter what the reaction was yesterday to some extent because, I mean, who knows, three weeks from now, right, just based on what the model can do or the model that's underlying it. I mean, it could pull ahead of Muse very quickly in this race, even if it's not in September 2026, right?

5:29Well, here's the thing. We have to figure out, is OpenAI going after consumers or if OpenAI is really going after enterprise? And that's a question they've got to answer internally. They always say they want to be an enterprise, but the behaviors and actions remind us of the early days of Google Cloud where it's kind of consumery, but it's really not. It's like a pro-consumer, a prosumer. and it feels like OpenAI is still in the prosumer stance and not getting to enterprise. Whereas Meta, you know that's consumer, right? That is completely consumer. It's going to be set up and made easy for the consumer to use.

6:01So I think they're kind of in prosumer phase and they'll make their way to enterprise or figure out we're going to do both and pivot that way. Right, and so yesterday really, I mean, I see what you're saying. It really felt like a consumer event in some ways still. It still felt like a consumer event. I mean, if it was enterprise, right, you have product roadmaps. You'd say, hey, in this timeline, we're going to build out these features. The partners would be announcing their piece of what they're doing on specialist thoughts. So the partners would be up there on force. You know, you've had enterprise CIOs come up and say, hey, here's what we've been building.

6:33Here's what's in the beta. This felt very prosumer. So let me ask you this then. If Google Cloud is the analog that you're using for a company that came out with a product that felt prosumer or geared towards maybe smaller businesses initially, I mean, it seems to have made the transition here. It's growing pretty quickly among enterprises. What worked for Google Cloud that you think OpenAI could employ to navigate this shift to enterprise looking ahead? It was an individual. Amos Kurian came and made Google Cloud more enterprise. Before Diane Green, they were prosumer. Everybody on the team said they understood enterprise, but they didn't.

7:17It takes a enterprise veteran to actually pull it off. And you've noticed there have been some changes at OpenAI as they're trying to get more enterprise veterans in place. And I think that's going to be the difference. Because at the end of the day, the customer is a CIO in a company, and they're betting your career on you. And if you don't have all the elements in place from having architects, having support, having the folks that are there to actually share with the roadmaps, collaborating on product design, you're not going to get there. And that's what the difference is between full enterprise, prosumer, and going to consumer.

7:49Well, and the other thing I've been thinking about is the security is of paramount importance to enterprises. How much was that a theme in yesterday's Dev Day? and how much are you seeing it be a theme in the products that they release? I think security is one of the things that people are assuming is going to be security by design. I mean, you can emphasize it and go down the line, but when you're seen as the one breaking security models, it makes it kind of funny. They kind of hinted at the OpenAI hugging face hack when that was about it. But I think in the audience, everybody assumes that, yeah, the security is going to be here.

8:28Yeah, I mean, but that's an assumption, right? Like, in this era, that is a significant assumption because I feel as though if they, I almost feel like if they didn't go over the top with the security message, I'm thinking of Apple's keynotes, right? I mean, Apple's keynotes, security, it's not just a line item on it or it's not an assumption. I mean, they make a whole big deal about this. They dedicate entire portions of the keynote to it. And so I almost wonder if that's the approach that OpenAI needs to go with all of their keynotes looking ahead to really hit home to enterprises. Well, I mean, Sam did mention that it was important to hold off releases until security was important.

9:09And I think that was important. That's one of the things that they're talking about, delays because of security. And so you're going to see more and more of that in terms of the conversation. But it's not just that, right? Apple is a different animal. If we're talking about dynamic security, dynamic security capabilities, there's only one company in the world that has that, and it's Apple. Right. But, you know, I am thinking about, like, you know, this is the part of the Apple keynote that, you know, frankly, when I'm watching it, I kind of glaze over because, you know, the virtual, whatever, virtual private cloud.

9:43Is that what it is? private cloud. Like, it's all the stuff that I don't care to hear about. And yet it's probably all the stuff that CIOs or, you know, anyone buying Apple for, you know, the decision between PCs or Macs for their enterprise use cases, that's the stuff they really care about. And I feel as though OpenAI needs to go over the top with that. I think OpenAI, Anthropic, everyone has to go over the top now. I mean, it's a different day. We've got Y2Q, we've got Quantum Day, Q Day, depending what you want to call it coming. And so all these encryption keys are about to be broken. All these people have been sitting on the sidelines grabbing passwords and information for that day.

10:23I mean, it's going to be huge. And so I think people are preparing for that. But I do agree with you. That messaging has to come clear. And Ray, before you go, I do want to ask you a quick question because you study the market at large. So we've seen this week, I mean, a number of IPOs have been delayed. It's hard to talk about OpenAI without talking. mean it well postponed is that what you're saying even shelved yeah shelved fine yeah although you know i uh you know i'm i still remember when you know the instacart era of uh ipos were shell they ended up coming out but my question for you is if the ipo market delays the anthropic ipo what impact do you think that has on the on the ai story at large i think it's going to be a game of efficiency, making sure you have enough cash on hand.

11:14And I think that's what's host, that's going to be the challenge. Can you grow your revenues fast enough? On the Anthropik side, they've shown they've been able to ramp up enterprise growth. And I think that's a big piece of it. And this is what I'm saying. It's really important. You have to figure out, are you consumer? Are you prosumer? Are you enterprise? Because you have to double down and take that market because the other guys are not going to. And this is why that extreme focus is important. If Anthropic does not go public by the end of this year, all hells are going to break loose in AI financing.

11:44Great. Well, Bray, I want to thank you for coming on. That is Ray Wong, the CEO of Constellation Research here on TI-TV. Ever since Google shook up the Internet with its AI-powered summaries, publishers have complained that the feature has dramatically reduced the traffic that Google sends to their websites. Now, the information has exclusive reporting that Google is paying some publishers for content that helps generate those AI-powered answers. For more on this, I want to bring on Alix Couture, who reported that story. Alix, welcome back to the show. It's great to have you here. Hi, Akash. It's great to be here.

12:18So, Alix, how much has Google's AI results affected the traffic that publishers have seen? Let's start there. Well, it affected publishers' traffic a lot, and it was quite a big disaster for publishers. Outlets such as the Washington Post have seen its traffic drop by 50 % in the past three years. World History Encyclopedia, which is a little smaller, has seen its traffic dropped by 50 % in the past two years. And more broadly, all publishers across the web have seen their traffic down by about 50%. So the thing is that by providing summaries in AI overviews, it basically reduces the clicks of the users on the websites.

13:15So users don't go to the websites anymore. And that's really, that really hurts the publishers' traffic. And so traditionally, has Google been paying anything to these publishers for using their content at all? Well, not really. It hasn't been paying content. And that's the big problem. So it did strike a lot of deals with different publications and different publishers to use their content. But it's only been small, not meaningless, but small deals, pilot tests or programs. But overall, it hasn't been paying publishers. And that's why the new pilot test that it just launched might be a little shift.

14:10And so tell me about this pilot test. I mean, what types of publishers do we know? Who they're working with specifically? Do we know how much they're going to pay for this pilot? Walk us through what you found. So basically, it started this program about a year ago, less than a year ago. And it approached a lot of publishers. And the idea of the test is to basically calculate the value of the content provided by those publishers. And then based on the value and how it helps contribute to its answers, it will compensate the publishers. So Google is trying to figure out a way to resolve that math problem.

14:58It's a mathematical problem. How to compensate fairly all publishers based on their contribution to AI answers. And so they launched that program with, now they're working with a little less than 100 publishers, roughly 100 publishers. And it's mostly small to mid-sized publishers. Why? Because the bigger publishers don't really want to be part of this program. because they're all they're all they're all suing Google for their data right yeah a lot of suing Google and so they don't want to be part of a program that's not gonna compensate them a lot because the payments are very like they're meaningless so for those big publishers they don't want to be part of it because they hope to basically pressure Google to do something better and more meaningful.

15:54So that's why they're not participating. However, the smaller publishers, their content is going to be used anyway. So I think they're not losing anything. They're just getting a little bit of money for the content they provide Google. Well, so let's fast forward a little bit here. So, I mean, this pilot program will happen. Google will make a decision about whether or not it's happy with it, whether the publishers are happy with it. I mean, I just want to get your sense for how much money we're talking about here relative to the revenue that these publishers would have generated from traffic previously, before the AI overviews.

16:36I mean, I guess the question is, is this at all making a dent for the publishers in terms of the traffic that they've lost? Or are we talking about peanuts here? Like, what's the scale that we're talking about? Well, that's the thing. It's peanuts. The payments are really meaningless, and most publishers are not doing it for the money. So, for example, one publisher that started the program a few months ago has earned 50 to 60K, which is a very small slice of its overall revenues. So it's really meaningless. And another publisher, okay, for other publishers, it can go up to millions. So that's a little more significant.

17:22But from my understanding, it's not common. It doesn't happen a lot. And it depends on the size of the publication. It depends on how much content it makes. It depends on a lot of things. That's why payments vary a lot. But for most small publishers, it's going to be less than$1 ,000 in a couple of months. So it's really meaningless. So most publishers, in fact, are not doing it for the money. They're really doing it because they want to get insights into Google's algorithm and And really into how Google works and what kind of content it's going to push and what kind of content it values, et cetera, et cetera.

18:12So they want to understand how Google's black box really works. And so far, they didn't get a lot from it. So they have calls with Google weekly, and yet they still have no idea how they're being compensated? Why? What does meaningfully contribute to answers mean for Google? Because that's the big idea of the program. It's to value content that meaningfully contributes to its answers. They don't even know what that is. Right. To a certain extent, it could just be a check saying, hey, thank you for enrolling. Here's your check for 60K. I do think that Google is working on it really yeah really wants to be able to um to to find a way to actually compensate these publishers uh fairly but i think google is struggling a little bit and that's why publishers still don't really know uh how they're being compensated at least what do you think this means for the fate of all those big lawsuits that we've seen publishers file against Google and other AI companies?

19:33I mean, will this have any impact on that? Will it mollify any of the concerns that these publishers have at all? I don't think so, because those big publishers are not part of the program. So it might, in the long run, if it really signals a shift from Google, I mean, Google will start really caring about paying publishers. And if it really signals a shift in Google's mindset, then maybe, maybe it will help. And it might, yeah, publishers might be happier with the pilot if they decide to be part of it at one point. But so far, it's so small and it doesn't include the big publishers. So for most big publishers, it doesn't really change anything.

20:32Yeah, they're still going to be aggressive with it. Yeah, it's one of those many programs that are not meaningful and that are just symbolic. But do they really mean something? We don't know yet. So we'll see how it plays out. Great. Well, Alix, I want to thank you for coming on. That is Alix Couture, our reporter here at The Information. Elise AI, a company bringing AI to property management and healthcare, raised$350 million at a$4 billion valuation. I want to bring on co-founder and CEO Minna Song for a conversation about the company. Minna, welcome to the show. It's great to have you here.

21:12Thanks for having me. So when I go to the website, there is a lot going on. You guys are building a lot. Walk us through exactly what the company does today. Yeah, so Elise builds Agentic AI for the two industries, housing and healthcare. And we do this to help businesses operate more efficiently by automating a lot of the day-to-day workflows that people have, all the routine tasks. Really things that can be done by automation and AI. So you started with property management, is that right? So this is like, this is if I'm a property, a leasing agent or something for an apartment building, like all the communications, rental agreements, that type of stuff is where you started?

21:56Yeah, exactly. So we started with helping people rent apartments. A lot of the things that leasing agents do is highly repetitive and mundane tasks. And now we want them to spend a lot more time with actual residents doing really high value things. And then we now have moved into all of the workflows that happen at a property. So people can really be elevated in their roles. What's one thing that your platform can do maybe that people might not expect? I mean, I'm a tenant. I'm sure there's work going on behind the scenes. Are there nitty gritty things that I don't even know are happening behind the books?

22:34Oh, yeah. These businesses are extremely complex. So you have to consider all the kind of protocols. They're both highly regulated industries. So you can imagine being able to handle compliance paperwork, for example. Lots of the nice things about AI is you can do this and help a lot of different people, maybe non-English speakers, get through really complex navigation, navigate really complex application processes that change by every jurisdiction. So it's really not simple to automate these businesses. All properties sort of act and have different protocols and policies similar to health care.

23:17There's a lot of nuance in how physicians run their practices, their scheduling, insurance, things like that. So why did you decide to go to healthcare next? I mean, these are two, like you said, highly regulated verticals, but I'm sort of trying to think of these similarities in the customer base and, you know, short of renting out office space for your practice. They don't seem that similar. So why did you decide to go there? Yeah, they're not similar from a business perspective, but the same mission that drove us into housing, which is improving life's most critical areas through AI automation, is exactly the same thing that we're doing in healthcare.

23:59So from a technical perspective, the core infrastructure is really transferable. So we're doing AI orchestration, workflow, automation, we're integrating with existing systems, and really we wanted to kind of build and take more advantage of the platform and infrastructure that we have built and apply it to an industry that had very, very similar challenges and really something that was an essential need for society. Now, you guys crossed$200 million in ARR earlier this summer. what proportion of that is coming from the property management business versus the healthcare business right now? Yeah.

24:41So our initial market and foundation of the company was housing, but we're really doubling down on healthcare. So, you know, for our customers, it's one of the system's most expensive pain points that we're tackling, front desk and call center operations, handling 90 % of patient conversations for our healthcare customers, handling call abandon, improving call abandonment is really, really beneficial for their business. So the size of our healthcare business is growing. Which one do you think will be a bigger proportion of your business? I mean, it sounds like, so we're not sharing the proportion of where the revenue comes from today, but which one is going to be a bigger business for you two years down the line?

25:27do you think? Yeah, I think they could be both be really, really massive businesses. I think housing is sort of a couple steps ahead of healthcare, but with how fast we're building healthcare products are, and the healthcare industry, quite honestly, is catching up because they have such a great need. Yeah. I feel as though, you know, I'm just thinking about the the folks that we've had on the show and where I've seen money being raised, I feel as though healthcare is a more competitive vertical for these backend automation processes compared to property management. I mean, that seems to be the case, right?

26:09Yeah, I think a lot of great companies are tackling the problem in both industries. I think there is a lot of competition in particularly hospital systems, where a lot of where we focus is on more specialty practices, women's health, dermatology, ophthalmology, for example. And I think actually a ton of value will accrue in these practices. And there is actually a lot less competition and people worrying about these really important needs. And so what do you think then is your advantage, not just really against those other specialty healthcare companies or the hospital systems, but I mean, I'm thinking about the big AI labs or the bigger companies expanding their footing into this space.

26:57What is your advantage to any of those companies coming into it? Yeah, I think it's speed. They have a lot of things to go after. And I do think in healthcare, they're working on more of the larger hospital systems. And our focus has always been how deep we go in our customers' businesses. And it means really understanding every little nuance of each specialty, which really changes. And I think it makes a lot more sense for them and what they've stated that they'll do is focus more horizontally. And oftentimes these customers we serve are, they want something out of the box that just works and knows their businesses before they even have to log in.

27:42Right. You know, I want to ask you about trust and earning trust from your customers, certainly as it relates to agents, because we're now in this era where we are asking agents to do more things for us, or at least that's what they're advertising to do. So given how regulated property management and healthcare is, can you talk about some of the ways that have worked to earn the trust of your customers to basically hand off tasks to them and allow them to do things for you? And then maybe what hasn't too, because this is kind of an interesting thing where just because the model gets better, it doesn't mean that I, a healthcare clinic manager, will allow the agent to control more, right?

28:29Yeah. Yeah. I think that's a great question. Look, I think it is about trust. There is a big need. So people are going to move in this direction over time. And trust building is really about is delivering value and in a way that's safe and for their businesses and for their patients and for their residents. I think it's really easy to put together a very impressive demo, but it's very hard to make something work in real life. And people are going to be sold a lot less through demos, I think, and a lot more through case studies and references. So we work on just building a great business and a great product for our customers, not trying to just show something really shiny.