Lux Capital's Deena Shakir on AI in healthcare: scribes are crowded, clinical trials and drug development are the real opportunity

help? You need like a world champion um like speedrunner or something like that. Uh anyway, we have Dena in the reream waiting room. We have Dena in the reream waiting room. Welcome to the stream, Dina. How you doing? Hello. So good to see you guys. Thanks so much for joining. Um sorry for the delay.

um why don't you kick us off with an introduction on yourself and I'd love to know kind of how you fit into the Lux Capital structure ecosystem, what you're focused on and kind of and then we can go into a bunch of questions from there. Awesome. Well, I'm Dena. I am a partner at Lux Capital.

We're a multi-stage venture firm. We love turning science fiction into fact. I know you've had Josh on here before and heard all about that. Um we do everything from preede to preipo, so truly multi-stage.

Um, and I have a particular propensity for healthcare, which I know we're going to be talking about today, but we are all very much generalists. Okay. Uh, let's just we'll we'll get into the other question later. I already have completely new questions.

Um, Delian aspiruhov over at Founders Fund, buddy of ours, was saying that the seed ecosystem is just kind of been eroded by founders just skipping rounds.

And it feels like that's a lot of a kind of a it's in large part because these like AI companies were able to justify, yeah, I need a billion dollars to train a foundation model. And so even if you're not training a foundation model, the multi-stage firms are like, yeah, we can do a hundred million dollar seed round.

No problem. Like you have a good pedigree, maybe there'll be an aquahigh or something. Um, where is the preede market? Where is the pre the seed market still interesting or do you disagree with that entire take and you think it's never been better?

You know, I think it's a tale of two worlds and honestly the nomenclature for rounds doesn't really mean much these days anyway. On the one hand, the advent of AI has made it more cost-effective than ever to start a company.

And you are not no longer judged in terms of growth by how many people you're hiring or how much money you've raised, but actually how much how far you can get with the least amount possible. And that is only going to continue to happen uh as we see the advent of sort of AI software tooling and and all of that.

On the other hand, to your point, if you are raising uh for a company that is infrastructure intensive, that requires high amounts of compute, where you're building your own models, and let's be honest, how many of those do we need? Those types of companies can and will raise$und00 million plus seed rounds.

So, you know, I think yes is the answer. It's never been better, but also you will very much see a tale of two worlds. And, you know, on our end, we're funding uh you know, tiny seed rounds and very large seed rounds. really depends on the company and what's needed. Okay.

Related to that, give me your take on this post from Caitlyn Bolnik Relos. Uh, you must be one of four things to raise right now. An AI high-f flyier, a seed company with the promise of being an AI high-flying company. Extra points if you're young.

An American dynamism reindustrialization company or biology enabled by AI. What do you think? Caitlyn always has the great hot takes. I think that's accurate. Although honestly it's not the best time for companies that are raising in the biotech space.

So I'm not even sure that I would necessarily include that in the list. But definitely every company right now needs to have an answer to what their AI play is for sure.

That being said, you know, if you're seeing Groundhog Day of pitches that are, you know, AI notetaker, AI agent, like yes, we all use these companies and they're wonderful tools, but what we look for is sort of the the operating system, right? the the actual picks and shovels, the really generational companies.

Um, and where the the innovation is not just the act of using AI, but something that's actually defensible and transformative. Yeah. Elad Gil had a post yesterday. He said, "AI markets crystallize for a subset of AI markets. It is suddenly clear who the finalists or winners will be.

" And so, he breaks down a few categories: LLMs, code, legal, medical scribing, customer service, and search and IR. How are you how are you thinking about AI markets within healthcare?

There's obviously a bunch of different subcategories, but at the same time, some of the biggest headline fund raises have been around these sort of medical scribe uh products or at least products with that as kind of the core wedge.

Yeah, you know, I I we've actually invested with Eli in a company in the health AI space which I'll chat a bit about called Blueote. But um yes, AI scribes I think were the lowest hanging fruit in healthcare. the you know one of the biggest pain points was with you know documentation and note-taking for physicians.

Um these companies were not immediately up and to the right. Many of them have been around for quite some time. Um but of course the advent of LLMs and and really the the the openness to adoption which has always been the biggest sticky point in healthcare has made them particularly effective now.

But there's still so much latency in healthcare. We are still talking about AI in the context of fax machines.

So there is a lot of opportunity outside of that and I get excited about uh you know the clinical research side of things the opportunity in clinical trials you know if you're talking outside of patient care we're in a company called trial library which is um you know using AI and and technology to rapidly accelerate getting patients into trial trials uh for life-saving therapies and frankly this is where pharma companies are willing to spend a lot of money to find those hard-to-reach patients companies like which we invested in with Eli lab, for example, are using generative AI to help get drugs to market, so to help with regulatory and medical writing, etc.

Um, there's a massive opportunity outside of scribes. And although scribes are very interesting, it is increasingly a very crowded market. Last count, I think it was over a hundred. Um, and there's probably so many. Wow. Have you seen anyone? I'm sure I'm sure more are being formed as we speak.

As we speak, people are reading the the Tech Crunch headlines and they're saying, "Hey, maybe I should create Seems like Stripe Atlas has gone down under all the pressure. Yeah, just kidding.

Um uh yeah on the note on the note though on the last company I mean I think one question uh that that's come up on the show is can uh what will it take to actually reduce the cost to create and bring a new drug to market and uh discovering the drug itself is one thing but then is there any hope of of reducing the cost of of of the next stages in terms of outside of the the paperwork and and things like that but It seems like if you have this huge kind of variable cost of of trials, uh is is there any is there any sort of hope of of bringing that down over time?

Yeah, for sure. I mean, the the sort of drug discovery piece of it was really, you know, V1 of AI in in this in this field and we've invested in companies in that space. Um and there are new companies that are forming in that space. But outside of that, there's a lot as you pointed out clinical trials, right? big area.

There's, you know, there have been lots of companies trying to apply technology there. Yet, still a massive pain point, particularly if you think about where the bulk of the spend is in pharma in specialty drugs.

The these are tens of billions of dollars of of market opportunity here to help get these drugs to market faster. And in order to do so, it often requires finding really difficult to reach patients. There's only so much right now that AI can do to make that happen. And so it requires really unique business models.

And the company I mentioned, Trial Library, is a great example of a company that is using technology, but ultimately to find and match humans with one another. And those humans are patients and those humans are providers. Nice. I want to go Shark Tank mode for a second. I got a pitch for a hypothetical company.

I'm not actually building this, but uh today uh there was a piece in the Wall Street Journal about Amazon buying a bracelet for $50 a pop. You get uh transcription of everything you say. It was about the company. They just buy a bracelet. They bought the company. They bought the company.

Have any of the AI scribes in healthcare uh done something like, hey, the doctor wears a bracelet and then the the scribe is, you know, there's a hardware play or there's a a wearable play. Have you seen anything like that? Is there an opportunity? I don't know if I've seen a bracelet per se.

There are over a hundred, so I wouldn't be surprised if one of them involves a bracelet. But yes, I have seen hardware plays.

the sort of idea of ambient you know not taking that um I've seen that in the form of cameras that are sort of you know adding on the computer vision element I've seen that of course in terms of you know different types of of tabletop devices there is an extra added uh concern with healthcare which is you know the elephant in the room and that is around PHI and privacy and so and often times that's overengineering it really what you need is to understand what's happening in that encounter with the patient and as it turns out you don't nec necessarily need, you know, a ring, a bracelet, you know, a camera hanging over you.

Often times audio and in many cases just uh, you know, AI enhancing text inputs is all you need to get to the outcome that's desired. And for that reason, you're out. Um, how are you how are you uh approaching this uh brain computer interfaces as a category?

There's been um Neuralink Neuralink Nudge announced a new round yesterday, but how how are you thinking about the category broadly?

feels like something that right now it seems like a very small group of exceptionally connected and talented teams are sucking up the vast majority of capital which I guess happens in in every market but maybe more uh pronounced in in the BCI space but do you expect that to be a category where a few years from now there's hundreds of of players or more concentrated until um people can really kind of derisk we talked about a teal fellowship company that was doing that kind of the other side of the barbell there But yeah.

Yeah. I mean, as I think as a trend at Lux, we were investing in the space before it was obvious. We were early and proud investors in control labs, um, which you may be familiar with, which was acquired. Yeah. Yeah. Rearen, the founder, is actually now a venture partner at Lux.

It's fun to jam with him on these topics all the time. Um, I my father happens to be a neurossychiatrist, so I have a particular interest in this space. He was really big on TMS, and I think that there's still so much opportunity there. We are uh actively looking um and spending time with a lot of these companies.

To your question, I don't know if there's going to be a hundred of them in the future. These are still very capital intensive. There's still a lot of science um that does need to happen. We're seeing companies at the earliest stages through our Lux Labs initiative.

Um researchers who are rapidly accelerating um on the kind of innovation side and ultimately to get to commercialization. that piece of it is the piece that I think is still uh still remains to be proven. Um but we are we're excited about what's to come in in healthcare broadly.

Um I'm I'm sure there's tons of opportunities for startups to build new companies that are kind of built on the back of the AI trend. Are there any places where you think that AI in a narrow sense would be a sustaining innovation for the existing large players?

I'm thinking of like in the in the enterprise it feels like you know Google might be facing disruption in search to some degree but like the Gemini API seems to be doing really really well and like they were set up well to serve you know a frontier class model as an API um and so they're leveraging that and they will continue and and Azure has seen a similar thing with Microsoft they've they've been beneficiaries and in many ways it's been a sustaining innovation for them um in terms of cloud uh but but what is the shape of like the big players in health care.

Where are they going to like just completely win? So, you probably want to stay away from it as a as a startup folk. I I love this question. I don't know if you guys know this, but I I actually got my start in healthcare at Google, helping to build what eventually became Google Health.

Uh back then, one of the first products that I got to work on was this knowledge graph for health, which was, you know, taking symptoms and conditions, mapping them together. So, if you searched for itchy eyes, you'd have this graph pop up that explains to you might have conjunctivitis. very novel at the time.

We launched Google's first HIPACO compliant product. It ultimately led me to venture because it became abundantly clear to me that the real innovation in these really intractable industries was actually coming from early stage startups. This was before the advent of LLMs and all of that.

I ultimately think that these big tech players are obviously going to be key distribution partners. Google is still the digital front door for a lot of healthcare. Increasingly, OpenAI and others are are becoming that.

But it is still where so many healthc care encounters begin before a patient even knows that they're a patient. Um that being said, do I think that the, you know, the top innovative companies in the healthcare space are going to emerge from these big tech companies? Unlikely.

But I think that they will thrive and exist in partnership with them. And the hyperscalers when it comes to AI are always going to play critical roles. Yeah. What what what role are the like the big tech equivalents in healthcare playing? Like is this the big insurance networks, big big hospital networks?

I mean, we've heard General Catalyst is buying a hospital. I I don't fully understand the the thesis there. Haven't fully dug in, but um but what uh what are the bigger players doing um that that might uh where they might just be just default beneficiaries of new techn?

I mean I've so I spent a lot of time speaking with the you know the leaders of some of these companies obviously they're critical important uh partners to our companies um and they are rapidly looking at solutions in AI so they were they used to be bottlenecks these are the distribution partners you know selling into health systems has always been a slog ask anyone who's done it they've got the gray hairs to prove it um I was in a board meeting recently actually with a major executive from one of the biggest national payers out there who blew my mind talking about how they were thinking about adopting clinical AI.

We're talking actual bots that are that are that are engaging with patients in a way that was far more ambitious than even some of the most you know out there VCs that I had seen.

If they're doing it that fast like we need to get on this quickly, it is clear that you know in healthcare you want to meet patients where they are and this is where patients are. They're using these tools constantly. Uh providers as well, there's a new generation of providers who expect more digitally.

Um, and as you can imagine, and you see this in pretty much every major earnings call from these payers, this is no longer just a nice to have, it's a mandate. Yeah. What about uh robotics uh in uh surgery rooms? Uh we've seen Yeah.

I mean, Neurolink specifically is has talked about building basically their own system in order to do these surgeries. That's like a very Yeah. No, they built it. They're using it very kind of like narrow use case for something that they're doing.

But how far out do you think we are from robots conducting kind of routine surgeries? Let's say you you um need to get an organ removed or something like that. Is that something that you'd bet on a human uh still doing 10 years, 15 years, 20 years from now?

uh or or are we going to be at a point where a human is sort of uh kind of observing a robot, you know, working but not necessarily like doing the actions themselves? I mean, it's already happening. So, we're not talking about 10 years from now. It's already happening across the board.

And robotic surgery is an area that we've invested quite a bit in. You may have heard of a company called Orus, which was acquired by Johnson and Johnson, uh in which we were lead investors. Um it's happening in other procedures, in ultrasound, um in IVF.

We have a company um that's on the software side but enabling computer vision through IVF. Um if you've ever had I don't know if any of you have had surgery but many many procedures that used to be highly invasive are now done laparoscopically through robotic surgery.

So it's one of the areas in healthcare actually that has seen the most adoption and it's still human assisted in in many cases. A lot of that ultimately comes down to um you know to patient preference as well as liability.

But I would not be surprised if we increasingly move more and more to a world um with uh you know robots at the forefront of surgery. What was your take on Fiji Simo's post kind of outlining uh therapy potentially is one vector that open AAI's chatbt core product could be impactful. That feels like an example.

Health was the other big category that she laid out as well. Yeah, it feels like meeting customers where they are and yet all of a sudden we're talking about, you know, a a research nonprofit is now competing in the healthcare space.

It feels like it's a it's it's definitely like an unexpected turn of events if you were looking at open AI as a health investment a decade ago. Yeah, totally. Well, Fiji's amazing. She's become a friend and we actually met in the context of healthcare. So, I know it's an area she's deeply passionate about.

It's not surprising.

It is wild but it's not surprising again meeting patients where they are and particularly for um you know a chatbot that does tend to heir toward you know psychopansancy that's constantly telling you great idea let's elaborate on that more etc like chat GPT in an era with an epidemic of loneliness can become a very close friend it can in many case diagnose and you know there have been some really interesting studies published in JAMAMA actually looking at the empathy factor comparing AI to humans and sometimes the AI AI actually does score better.

Now, do do I believe that therapists will be replaced by AI in the near future? No. Um, but the fact of the matter is people are already using it. The numbers are staggering. They're using it for this case now. And so, I think what we need to think about is what are the guard rails?

How do we ensure that that, you know, you don't end up with the types of situations that have made headlines, you know, whether it's suicides and so on. Um, and and how do you ensure that this is done responsibly?

Um, and I'm excited that someone like Fiji is at the forefront of that because I believe that they are really thinking about um not only the the ethics but the responsibility of having human lives in their hands. Totally. Yeah.

It feels like something that's going to be like back and forth in the news cycle constantly but ultimately, you know, top of mind and they have all the capabilities and teams in place to to, you know, execute effectively in that category. Um, but it's gonna be it's gonna be a lot of PR back and forth, I'm sure.

Anything else? No, this is great. This is great. Thanks so much for hopping on. Awesome. Great to chat with you guys. Take care. Have a good one. We'll talk to you soon. Um, let's go to this post from Nick Carter. Did you hear this? Did you see this? No. Gonna gonna own you. Oh, wait. No, you're a millennial, too.

Um, he says, "Millennial Millennials will be slash are the peak IQ generation. Zoomers or iPad kids labbotomized Gen Alpha will be largely incapable of learning anything due to AI. Tyler, what do you think? Well, Tyler's you're Gen Z, right? You're not Yeah, I'm Yeah, I'm definitely Gen Z.

Do you think you're an iPad kid? Do you think you're lobotomized? I think that's a little bit younger than me, but I think that's probably true. Like, how do you avoid just replace iPad with Tik Tok? I think like if you talk to the average like I mean look, I don't want to you know Yeah. Yeah.

You don't want to throw your whole generation under the bus. I think there's a big power law and you'll see it like grow more and more with Gen Alpha. What separates the high performing zoomers from the labbotomized brain rotted ones.

I mean I think it's you can probably just like if you need like one thing it's just like do they use Tik Tok daily? That's like you probably get like 90% of the the Yeah. I wonder like is it a screen time thing you have to cut it off early?

I don't think it's necessarily screen time cuz you see a lot of kids who just like if you're like reading like less wrong or something. If you ever talk to someone who like obviously like read a lot of like blog like online blogs like that, they're like super cracked. So it's not necessarily screen time.

Maybe one one thing is I I feel like maybe the the kids that have it the worst are like 10 years old right now where I feel like parents today there's a general awareness that having your kid glued to an iPad is not great.

If you see I'll be at a restaurant and I'll see a kid just like drooling over their iPad for two and a half hours not engaging with the parents or whoever they're with. Uh I I think that's like fairly frowned upon now. Like I I I got an iPad uh for my three-year-old, but he's only used it on flights. Yeah.

Like he he just doesn't he only thinks of it as a Yeah. flight device. And that's purely because I don't want other people to be angry if he's going to, you know, cry for 4 hours. Tyler's over there being like, I'm not addicted to Apple products, but I'll do anything for an iPhone. Which one is it, Tyler?

Yeah, you got to pick. Got to pick. Any more progress? Uh, I'm still working through. I'm going like very slowly. Okay. So, yeah, because there's no time, so you're just really I can't see my score like until the very end. So, I think I'm doing better, but it's very slow.

And are you like actively on a run in ArcGIV3 right now? How do you what do you mean? Like like do you kick off a new new test or can you optimize each one individually? Like I think you can do each game individually. But you can't do a specific level.

Like you have to do the whole thing, you know, one through eight or whatever. Oh, okay. Okay. Got it. Cool. Interesting. I need I still need to do all three because I've only done the first one. I I actually haven't beaten it. Um Did you time yourself? Well, no, because I did it in parts, right?

I did like a couple minutes before the stream and then I did five minutes on the stream and then you know I I I I think I I think I know enough at least to be like to probably have an