The Pulse by Wharton Digital Health
00:33:14 · Arpan Parikh
[00:33:14] S1: Yeah. It's a great question. So Kyle, who is our cofounder and CTO, is a is a genius, and and he's architected things in in very intelligent ways. I'll I'll talk about this from two angles, like, from the product angle and then one from the business angle because I think they're both relevant. Right? From the product angle, in terms of the outputs we actually show to our users, we've architected every piece of that, you know, data ingestion and productionalization pathway to be agnostic to model. So we can, at the flip of a switch, like, point every step in the funnel to whatever model we want based on the flavor of the day, like, which model is giving us the best combination of quality of output, speed of output, and cost of output. Right? So, for the most part, so far in Kin's history... And, you know, Kin has been live as an app publicly for about eighteen weeks, so not not super long. Almost all of that or all of that has been contained inside GCP, inside Google. Right? And we made that decision for a couple of architecture reasons and security reasons. And we also, you know, in surveying the models, found that different iterations of, you know, Gemini models were were the best combination of cost, quality, and speed for us. So, you know, if you've used Kin or if any listeners have used Kin, like, you know, from a visit recording, there's multiple outputs we generate for our users. Right? We generate the title of the visit. We generate the clinical summary. We generate the detailed topics, and then we generate the action items. Like, we actually point each of those to different Gemini models because to generate the title of a visit, like, you don't need to use a frontier model for that. Right? But to generate the detailed summary, we wanna use a model that's a bit more robust. So we're very, like, picky and selective, and we have that flexibility in real time. Okay. So then flipping to the the business side, I personally can't imagine what it could have been like to run a business or be a CEO in a pre LLM era. Everyone says this. Right? With, like, whatever the preeminent technology of the day is, I couldn't imagine what it would have been like to do x without y. But we've been extremely intentional, and this is a combination of Kyle plus Amit in the way we wire up the whole business. And we literally wire every single system in in...
00:35:41 · Arpan Parikh
At Kin on the business side into data warehouse. On top of which we have a company harness, which Kyle has actually built from scratch. And what that lets us do is we have this, like, single source of truth, this single chatbot that knows everything about Kin in real time. And not only does it know everything about Kin, it can almost... It can do almost anything we need done for us inside this data warehouse. And that is incredibly powerful as a management team for a business to have your fingers, like, on top of and inside the data in a real time twenty four seven, three sixty five in a self serve way. So if I had one takeaway here, I actually think the business stuff is a little more interesting than, like, the product side stuff. And I would encourage folks who are building businesses or scaling businesses. Obviously, it's a privilege to have built a business natively this way, but if you're early in in the game, like, it's... It is a 110% worth the effort and intention to to wire it up this way because it gives... It will give you and it gives the management team a type of leverage that I don't think has ever been seen in, like, corporate America before.
00:36:55 · Chase
[00:36:55] S0: Yeah. Very cool. I know you mentioned AI, and and I I can't get through this podcast without asking Mhmm. Folks about AI. But how is... One thing about AI in particular that I'm curious about is how is Kin thinking about using AI for administrative purposes, like synthesizing notes, which it seems like you're already doing Yep. Versus sort of getting into the clinical territory of providing clinical guidance. There there is a fine line between, you know, nudging a patient along to take another action actually providing clinical guidance. So curious how Yeah, you think about
00:37:30 · Arpan Parikh
[00:37:30] S1: it's something we have a very strong POV on, and we've had this since we started the company, is we want to create a very strong line of distinction between, to your point, the nudges and the summarization and creating artifacts from a visit and actually making clinical recommendations or driving clinical outcomes for a patient or for a user. Like we don't even, they're not our patients, They're our users. And there's many, many reasons why that distinction is important. But the biggest is actually, I believe from a, it's a competitive and differentiating moat for us, which is most consumer health companies, or you could say almost every consumer health company that exists today, has this similar storyline. Right? And the storyline most consumer health companies weave is the existing healthcare system sucks and health insurance sucks. So come to us and we're going to be some combination of better, faster, and more convenient for you than the traditional healthcare ecosystem. Now I'm not gonna say I don't disagree that, generally speaking, the healthcare system and the health insurance system in this country are not ideal. Right? However, going back to my broken record example, the 90 year old in rural Ohio, like, he doesn't have the privilege going to function for his labs when his CKD is decompensating. Right? He has to go to the shitty lab corp around the corner to get his labs drawn. Mhmm. And the woman in Nebraska, like, she doesn't go to council health. And like, I have very good friends who are building and running council health when she's diagnosed with breast cancer, right? And she's going to University of Nebraska's medical center. So we have this strong belief that the majority of Americans, like, when they actually get sick, I. E, they actually need real help, like, they go and are going to continue for a long time going to the traditional health care system. And there are enough people trying to build these alternative pathways for care. What we wanna do is actually make smoother and more efficient and less painful all of the people who don't have the privilege of alternative pathways of care. Right? And so with that mindset, I don't want to or need to deliver care because I don't want or need to disintermediate the healthcare system.
00:39:55 · Arpan Parikh, Chase
What I want to do is turbocharge the doctor in the academic medical center or in the rural primary care office in the Upper Midwest. And I wanna draw a strong line between delivering care, which we don't do and I don't think we will do for a long time, if ever, and being the layer that helps the patient understand the care that their local on the ground licensed healthcare professional is trying to provide for them. [00:40:22] S0: Yeah, yeah, no, that makes perfect sense. Okay, two questions on more of the sort of strategy front for Kin. One thing I I've been thinking about is, I think AI scribes have been talked about for several years now, but always on the provider side. Right? So you have companies like a bridge and ambience, and then you have the the big tech players like Microsoft and Epic Yep. That have their own scribes. And so I'm curious what your thinking is in terms of your competitive moat. Because, hypothetically, these companies are also thinking about how can we equip patients with the same capabilities? So how do you think Kin's model is more defensible?
00:41:04 · Arpan Parikh
[00:41:04] S1: Great question. I think of it in three primary ways. So mechanism one is the motion of building a delightful and magical consumer product, which is what Kin is, is very different from building enterprise SaaS software. And I think of the clinical scribe and EHR companies as primarily enterprise b to b SaaS software companies. And if they've shown us anything without naming names, the largest EHR in the country and their patient facing products has a negative net promoter score. Right? So, like, they've proven that they don't have the capability of building beautiful, magical consumer experiences. On the other hand, we've built a product that even the 85 year old who's never used ChatGPT or or Claude absolutely loves. And when we interview our users, like, they tell us things like you will not pry Kin from, like, my dead hands. Like, that is the level of love they have for the product. So I think that's a big, really big piece. It's a special competency as a company to be able to build this type of product. Okay, number two is trust, right? There's something to Kin about being a business literally built by two double board certified practicing physicians for patients, right? And we've leaned into that because it's a very valid piece of our story in DNA, right? And when you think about tools and products that Americans use for their healthcare, trust is probably the most valuable currency, Right? And now think about the Frontier Labs, think about the electronic health records, and think about the clinical scribes. Either across all of those, there's... One of two things is true. Either there's neither a positive nor negative association from a trust perspective because, like, the consumer doesn't even know about them. Like, most Americans don't know what Heidi or a Bridge or Ambience or Freed are, honestly speaking. Right? Or for some of them, Americans probably have a net negative perspective from a... From the trust angle. Right? And and when it's negative for a massive behemoth company, it's really hard to, like, steer that ship in the opposite direction and even get it back to neutral, let alone, like, the super positive side we have it on. So trust is number two. And then number three is is distribution. Right?