Dreamforce 2025, AI Challenges in Healthcare, Pushback on AI Coding Tools | Oct 15, 2025

0:13Welcome, everyone, to the Informations TI TV. My name is Akash Pasricha. It is Wednesday, October 15th. We have got a great show planned for you today. We are talking about Salesforce's Dreamforce conference and the big announcements it made with openai. and with Anthropic. We're also talking about the future of health tech with Christina Farr from Second Opinion Media. We've got a great discussion planned for you about the rise of the cursor resistance and why some people in Silicon Valley are pushing back against using AI coding tools. And last but not least, we are talking to the CEO of Descartes.

0:46It is a great conversation. You do not want to miss it, but I want to start with Dreamforce. Salesforce's annual conference is making quite the stir as usual. One of the biggest bits of news out of the event with Salesforce's renewed emphasis on its relationship with OpenAI and also its relationship with Anthropic. I want to bring on Erin Wu, who has been at the conference all week, and Kevin McLaughlin, our enterprise software reporter, who has done a ton of great reporting on Salesforce. Erin and Kevin, welcome back to the show, I should say. It's great to have you both here. Thanks so much for having us.

1:19So Erin, let's talk about the vibe this week. What is Dreamforce 2025 all about in San Francisco? The first thing you have to know about Dreamforce is Dreamforce is absolutely massive. It's like 45 ,000 people downtown Moscone Center talking about enterprise software. And this year's main question, this year's main theme is agents, agentic AI, and are people actually using it? There's also just so much atmosphere and juxtaposition at this thing. There's a little bit of awkwardness with Marek Benioff last week gave an interview to the New York time in which he called for sending in the National Guard, but at the same time, he's starting this conference with a Hawaiian song.

1:59He's talking about setting aside our divisions and coming together as one Ohana. So there's a lot going on at Dreamforce this year. Okay. Kevin, talk about the announcements this week. We saw the news with OpenAI and Anthropic. I think there's a little bit of history here. Walk us through where that relationship has been traditionally and then where it stands right now between them. Well, I think you can take them both, Anthropic and OpenAI, together. Salesforce initially partnered with OpenAI in March of 2023, and it was a big deal because that was the first time that Salesforce was talking about integrating its products with the hottest large language model out there.

2:36And then, of course, Salesforce has subsequently partnered with Anthropic as well. So it has these relationships with cutting-edge technology that it can incorporate into its main AI product, AgentForce. I think the relaunched or the expanded partnerships that they announced yesterday, there's some new stuff in there. For example, OpenAI, you'll be able to access Salesforce CRM records from within ChatGPT and other Salesforce products as well. That's interesting. But I think on a higher level, it feels like Salesforce is basically telling customers we are still relevant in terms of having relationships with the top AI labs.

3:22And as a result of that, customers are going to benefit as each new advancement comes out. Salesforce customers are going to get those benefits. Did this not surprise you at all though, Kevin? Because we've talked a lot about the corporate data wars and Salesforce's sort of protective attitude towards data that it sits on data inside Slack, for example. I mean, the idea that OpenAI, that ChatGPT could access CRM records. I mean, it doesn't that sort of go against traditionally what their attitude has been towards applications like this? Yeah, I mean, I had the same question. In some ways, you could look at being able to access CRM from within chat, GPT as not a good thing for Salesforce because it might cause people to not use Salesforce applications as much.

4:06I don't really think that's the case at this point. I think that that use case, first of all, is right now a pretty niche one. I don't think it's something that's going to threaten Salesforce in the near term. More importantly, I think that, like I said before, I think, you know, OpenAI is, it's on the cutting edge and Salesforce wants to be on the cutting edge. And so you have to realize in the Salesforce universe, everything is about optics and messaging. And one more thing that I'll add about the conference this year is that last year, Mark Benioff's message was, go forth, build agents, it's straightforward, it's fairly easy to do.

4:46This year, I think there was a marked change. And in the Q &A, he was basically, someone asked him, like, you know, why aren't more people using Agent Force? And his response was that enterprises can't just change their processes overnight. This stuff is complex. It requires architectural changes. And so I thought that that was interesting to see Benioff sort of reflecting what a lot of people think is the more realistic view of agents is that they will take time to take root in the enterprise. Right. Aaron, what was the reaction of folks on the ground at the conference, the customers that are wandering all around San Francisco, the OpenAI and Enthropic announcements?

5:28Were they excited about it? Were they surprised? Were they already anticipating something like this? Yeah. So broadly, people are excited about AI, people are excited about agents, but it was a lot harder to find people who have actually started putting this stuff into practice. One of the other interesting things that Mark Benioff said in the Q &A was he was saying that this is a moment where technological innovation is actually ahead of customer adoption. And that's definitely what I found on the ground talking to people. So excitement around using this to, for example, automate customer support, follow up with customers, like summarize meetings.

6:06There's a lot of enthusiasm about this, but a lot of people who are still like, oh, like now I'm starting to use this. Like I turned it on a couple months ago. We're playing with different things. And so a lot of customers who haven't yet put this into production, and that broadly tracks with the numbers that we got out of Salesforce, which was that 12 ,000 companies are using this out of 150 ,000 companies were the numbers that the NEF gave at the keynote. And it's worth noting there that he's not even talking about how the extent to which this is put into production. He's not really making a difference between the idea of there's one agent that people are kind of testing with people's businesses run on this.

6:49And so even with that numbered analyst that I talked to was like, take that with a grain of salt. We don't know how much usage this actually means. Kevin, at the same time, Aaron has a good point. Are we hearing at least more stories from customers saying, hey, I have been able to use it at least for certain applications in ways they haven't talked about previously? Sure. We've written about the challenges Salesforce has had in getting adoption for Agent Force. In some cases, it's because customers just aren't ready and having their data ready for AI. In other cases, it's because the product itself has had some glitches and had some issues with inaccuracy.

7:27And also, I think Salesforce has kind of been talking about the same cohort of early customers until now. You're talking about home building company Lenar, education services company Pearson, OpenTable. I think this year we saw five real examples of Salesforce customers that are not ones people have heard about before who are using it in their businesses. And I think that there is real progress here based on what those customers talked about on stage, because they're not just using AI for automating customer service, which is basically the most sort of well-worn use case at this point. But for other things, like, for example, Dell, Michael Dell joined Mark Benioff during the keynote to talk about how Dell is using Agent Force to manage its supply chain.

8:17And that was really interesting. And then, of course, Salesforce launched an agent that does basically what ServiceNow does, which is to process IT tickets. And FedEx had someone up on stage as well talking about how they're using that particular product. So they are making progress. And I think it's just a matter of what Benioff said earlier. Enterprises adopt technology slow, but eventually they do get there. Right. Well, I think it's a fascinating time for Salesforce, and it's certainly an interesting time for this event to have been put on. Aaron, we'll let you get back to it. I know there's a lot more programming for you to get to.

8:56Kevin, thank you so much for helping us understand the context around it. That is Aaron and Kevin, two of our superstar reporters here at The Information. Okay. Healthcare is one sector Silicon Valley has been trying to innovate in for as long as technology has been around. It always feels like progress is slower than most people hope, but many people are still touting this moment to be different with AI being introduced into the conversation and also new focuses on topics like longevity. I want to bring on Christina Farr, a healthcare-focused investor and advisor. She is also the CEO of Second Opinion Media.

9:31She's got a whole host of things going on. Christina, it's so great to have you on the show. Welcome to TITV. Oh, it's so good to be here and just fun to get to do this with you, get to hang. I know. Yeah, gosh. Well, I mean, you're a vet, though. Back in your old broadcast journalism days, I mean, this was your daily cadence as well. So we're excited to have you on the show here, albeit in a different capacity. So let's talk about AI and healthcare broadly. You know, we've had people on the show and most of them have been entrepreneurs and the entrepreneurs have come out saying, look, it's just a matter of time, right?

10:06All non-touch aspects of healthcare will get replaced by AI at some point, you know, whether it's six months, 16 months. What I want to talk to you about is the doctor perspective on this, because we haven't yet had as many doctors on the show we should. You, of course, talk to a lot of doctors. And so when you talk to physicians, what is their reflection on the extent to which AI could replace healthcare? And do they actually believe it? I mean, walk us through that. Yeah, I think it's really changed where in previous kind of waves of technology coming into healthcare, you saw a lot of skepticism in the physician community.

10:41With AI, I actually see the opposite right now. A lot of them tend to be very early adopters and dabblers. Even the ones that you would expect to be very conservative seem to be on board with this. I think they really appreciate that it can do things that they really do not like doing in their jobs, including a lot of administration and note taking. AI is very good at manual, repetitive tasks. So I think that's been very key. And then, you know, recently I was on a panel or listening into a panel of a group of physicians and somebody made the comment that in the old days it used to be considered lazy if you were a physician that used one of these tools like UpToDate or Doximity to look something up as part of the way that you practice medicine.

11:30And nowadays, with just how intelligent AI has gotten and how much more knowledge AI has at its fingertips, it's lazy not to use AI in the context of patient care. So I think physicians have really come around to the idea that at the very least they should be leveraging this as part of practice and that there are things that, candidly, AI can do better than humans. Are they as excited on the diagnosis side as well? Do they think they'll be able to? I mean, I get the radiology angle to this, but are they saying that, you know, they will be able to diagnose things as well? I think it depends on how the information is presented to them.

12:12If they are being told by a tool, you know, you should do X, Y, or Z, I think there is some resentment towards something like that. But a lot of the companies have done it differently. You know, take, say, a Doximity or an Open Evidence where instead it's more of a set of options. Like, here is option one, here is option two, here is option three. This is what the evidence says. Pick amongst these various protocols. And I think they like having some degree of agency, but be guided by what the research says and what the AI comes up with. And that's where I'm seeing the sweet spot is, you know, you're not replacing them.

12:48You are actually guiding them, but you're presenting them with maybe information that they don't have because no human being can ingest, like, the world's knowledge. as much as they think they can, but it's just not realistic. And so I see where the AI and the human kind of can work together and collaborate as being the most effective. Now, one of the things you wrote in one of your columns that I found interesting is you wrote, candidly, one of my big fears is that the first wave of AI companies does very little to strip out unnecessary healthcare costs. So it sounds like you do have a bit of cautious optimism about the technology.

13:24Talk a little bit about where you do have big questions here about reducing costs. Yeah, I am worried about this. I think we sometimes talk about AI as if it's like automatically going to have this immediate impact of reducing costs in our system. But if the incentives right now are all set up to increase costs, which is why costs have been going up year after year, I think some of the near-term revenue generating opportunities for companies will really be to add cost. And it may be costs that, you know, one could argue is correct. For instance, if a patient and provider have an encounter and you use some kind of ambient scribing tool to document that encounter and then there is some amount of coding that comes out of that that is correct coding and maybe the physician was undercoding the encounter before.

14:11But still, that is going to add some level of cost to the system. So where I'm very interested right now is, is there anything that even in the longer term will strip out cost? Because right now we have literally hundreds of millions of dollars going into pure waste every single year. And to me, that is the sin of the healthcare industry is the waste. It's not necessarily that we just overspend on care. We certainly do that. But imagine what we could do with hundreds of millions more dollars at our disposal. And that, I think, is the exercise that I'm constantly working through in my own mind as I look at some of these new technologies.

14:51Right. The other interesting thing that you wrote about is this idea that there are sort of two distinct health systems in play right now. Now, one is the health system that is very much run by the insurance. It's covered by insurance, essentially. And the other is sort of the private health care sector, really, you know, these consumer technology devices that have come up that are helping us understand our own bodies better, our health span and, you know, the way that the longevity focused industry has developed. Do you think this is a good thing, a bad thing? I mean, how do you think about this broadly?

15:28I am a little worried about it. You know, I see patients getting, say, compounded GLP-1s for weight loss on the internet for a few hundred dollars a month. And that's free because it's a more affordable option. But their physician, the family physician, has no idea that they're taking these medications. And it's really important that that information be part of the medical record. But when you have this, what I call the tail of two systems, where we have this burgeoning cash pay ecosystem that isn't connected at all to the traditional healthcare system, then you could see just so many information gaps that could even lead to patient harm.

16:09If the physician doesn't know that a patient is taking hormones, doesn't know that a patient is taking peptides, and then prescribes them something that maybe would have a bad effect with these drugs and has no idea, no way of knowing that unless the patient has the agency to tell them directly, I can see things happening as a result of that that would be harmful to the patient. So I think we need to build some kind of bridge between these two systems because the cash market, in my mind, is not going anywhere. And the traditional healthcare market isn't going anywhere because people still need to use insurance for things at a very high cost.

16:46So if you assume that both will continue, then how do we then kind of form that middle layer between them? And I think there is an enormous opportunity there for companies. Any particular companies you're watching in this space building that opportunity? Funnily enough, one of the ones that I advise, a company called Vibrant Practice, is doing exactly that with building that kind of middle functionality between the longevity, functional medicine type clinics, and the traditional healthcare system. That's just one example that I've seen recently, but I think we'll see many, many, many more. And as I speak with health systems, they've echoed to me that this is something that they are very concerned about and that they, you know, in many ways are envious of this cash pay system that has emerged because they wish that they could capture the patients in the same way because it's turning out to be a front door for healthcare that they don't have.

17:40And so I think there's going to be some M &A opportunities, a lot of partnerships as well between these two systems going forward. Great. Well, Christina, I want to thank you for coming on the show. It's a fascinating discussion. And I will say, next time you come back on the show, bring the CEO of that portfolio company on with you, and we can have a discussion about how we can bridge this gap. Because I do agree, these two systems are sort of operating in their own separate lanes right now. And I do think it would be beneficial for them to talk to each other a little bit more. Christina, thank you so much for coming on the show.

18:12That is Christina Farr, the CEO of Second Opinion Media here on TITB. Okay, coding tools have been one of the most popular applications for AI, and it has meant venture funding has poured in for top companies like Cursor. But it turns out some people in Silicon Valley are holding out against these AI coding tools and even forming a resistance of source to using the technology at all. My colleague Rocket Drew wrote a great piece on this, and I want to bring him on to talk all about it. Rocket, welcome back to the show. It's great to have you thanks for having me so let's talk about the cursor resistance before we talk about it did you get a haircut by the way that's a that's a great haircut man last time your hair was way longer i know i know i'm saying this because it was long it's a big improvement i had a wedding it kicked me into gear i like it it's it's it's good um i i should also say we have some some photos of the show on our wall and and i have that photo of of that particular haircut on anyway Let's talk less about haircuts.

19:13Let's talk about coding. How did you get the idea for this story in the first place? Well, we'd been hearing rumblings from engineers who were a little disgruntled at how much they had to use these AI coding tools. I mean, by and large, the narrative has been that these tools are extremely useful to make engineers much more productive. And certainly that was the experience of many engineers we talked to. But we had a hunch that some engineers were not so happy about them, and we followed that lead. Okay. And so what did you find? Well, we found that, in fact, some engineers don't like that they have to use Cursor so much.

19:47It's Cursor, quad code, this genre of sort of AI coding tools, programming assistance. And there's sort of two buckets of reasons for that. Some engineers pointed to the technical limitations of the tools. The tools are just not good enough in their eyes to be useful to them today. At least that's what they said. And for other engineers, there's more sort of an existential angst that these tools raise. Because if the tools are good, they're good enough to do the engineer's job, then maybe your whole career starts to look kind of obsolete. What's your value anymore? What are you bringing to the table?

20:20So these two ideas seem to be kind of intention. Like either the tools are too bad to be useful or they're so good that they're going to take your jobs. But in fact, we heard both of these stories many times and sometimes from the same people. and you know one of the things that struck me about the story that you wrote is it's really the the really technical people that are sort of pushing back the idea that this that this technology cannot do uh it's not good enough for them i mean do you get the do they have any sort of confidence that the tools will get better or is it really just the the control and saying look i i just don't want to use a tool at all i want to do it myself yeah that's right i mean it was tricky to generalize about which groups are most reluctant to use these tools.

21:05But to the extent we can, it did seem like senior engineers in particular were reluctant because they feel like, well, one, they can just generate a thousand lines of code on their own. They can do that by hand. And in fact, they feel like they can do it better than the AI can do it. So there was definitely some resistance among senior engineers. I think it varied whether they see the tides changing. I mean, some of the senior engineers definitely realized that the tools are progressing very quickly. And the generation of AI coding tools that exists now is so much better than the tools that existed a year ago.

21:38So I think some of them do see that change is coming and know that at some point they'll have to adapt to them. I think others see these as sort of fundamental limitations of the tools. Like they're just not smart enough to handle the kind of complex logic that they engage with on a day-to-day basis and write the kind of hyper-efficient code that they're writing to make AI models run, for example. I will say that this story actually got a lot of comments on, and I think people are very passionate about it. And, you know, one of the people you mentioned in the story actually let go of engineers because they were unwilling to use these applications at all.

22:21I mean, this seems like something that people have very much drawn a line in the sand and they're actually willing to let go of engineers if they don't meet the philosophy that they've sort of developed. Yeah, that's right. And I want to comment on what you said before, that it seems to touch a nerve with people. Like this topic seems to be something that people are really passionate about. And my pet theory for why that is, is like you said, AI coding tools have really exploded in the past year. I think in some ways, AI coding has become a bit of a poster child for generative AI overall. Coding should be one of these tasks that generative AI is very well suited for.

23:00I mean, it's entirely text-based. Often you can give an AI tool a sort of self-contained, bite-sized problem to work on. There's plenty of data out there to train these models, and it should be relatively easy to check whether they did a good job or not. So AI coding seems like one of the perfect applications for AI, and we've seen the valuations of AI coding startups reflect that promise. So when engineers start to point out the limitations or to point out their sort of concerns with these tools, I think it gets right to the heart of people's anxieties and fears about generative AI as a whole industry.

23:35Right. Well, look, I certainly did not expect that so many people would be resisting against it. Of course, it's hard to know the exact number, but it's interesting to hear that there's anyone at all that is pushing back. Thank you, Rocket, for coming on the show and explaining it to us. It is a great piece. We will link it in the show notes. That is Rocket Drew, our AI and robotics reporter here at The Information. Okay, AI video generation has been a big conversation lately as OpenAI introduced the newest version of its Sora models on a social media platform to go with it. Many other companies are vying for a slice of that market as well.

24:12Descartes is another one of these startups. The company raised$100 million at a$3.1 billion valuation back in August. And I want to bring on the CEO, Dean Liedersdorf, to talk about this moment for AI Video and for his company. Dean, welcome to the show. It's great to have you. Thanks so much for having me. Great to see you this morning, Akash. You as well. So let's talk about what Descartes does and how it's different from all the other video generation platforms that we've seen come out recently. recently? So I think that video generation is really starting to have its breakout, especially as you mentioned with Sora 2 coming and standardizing this for the industry.

24:48We have been focused on video generation for the past two years. We recently launched our real-time video models, which are the only ones currently on the market. We launched a demo roughly five weeks ago. We've seen it take up in lots of different markets, whether it's camera apps, VR, social media, and gaming, I think that AI-generated video is really here to disrupt and change lots of different industries and we're really interested in that. We're going to see a lot more coming over this next year. What does that mean, real-time video generation? So, you know, we couldn't we wanted to share my screen in this interview, but we couldn't for technical reasons.

25:26Maybe next time we could. But if I could, what I'd show you is that I would turn into a dragon. I would hold up a pen and would suddenly turn into a sword. And this all sounds like magic, but you can actually do it today. Okay. If you go on our side, we have a bunch of products that we released playing around with this, whether it's in VR, whether it's on your phone, whether you're using it for Twitch streaming, or you're using it in order to play inside a game like Minecraft. Yeah. And it's like a regular video model, which, you know, Sora and others that you know, but just imagine that it works instantly.

25:59So, you know, you're sitting there with an orange shirt, I can say, hey, turn that orange shirt into a suit or take a red pen and change it into a sword but we can i mean we can already do some form of this already with tools like snapchat and you know i can i think i've done the whole um i don't know what is it called memoji on on on iMessage i can i can turn my face into an elephant and i can talk like i could do some of this already so i think it's exactly what we saw with regular video models that you could do lots of that before whether it's with tools like Adobe Studios with Photoshop and stuff or whether it's with simple tools like do specific stuff but let me ask you this have you ever seen a tool that actually turns you into say a new movie comes out, Elza and you actually look like Elza and you don't just have nice blonde hair and just a few touches of makeup upon your skin.