Navigating the AI Landscape Without Titles with Ankit Jain, CEO of Infinitus Systems

People thought we were doing black magic. They did not believe that a computer agent or a voice AI agent, as we now call them. have a conversation that was 30, 40, 50 minutes long. after December, 2022, when chat GPT became part of the global zeitgeist.

People believed that AI could do magic, and so we no longer had to sell magic. Healthcare is designed to be reactive because of a scarcity of resources. AI flips that on its head. We can now be proactive.

Once you create titles. starts becoming motivated by titles. If you are an engineer two, your goal is to get to engineer three. And once you have titles, it leads to entitlement because people start trying to assert power based on title and make decisions based on title.

Welcome to SaaS Scaling Secrets, the podcast that brings you the inside stories from the leaders of the best scale up. B2B SaaS companies. I'm your host, Dan Balcauski, founder of Product Tranquility. Today I'm excited to welcome Ankit Jain, co-founder and CEO of Infinitus Systems, healthcare's age, agentic communications platform.

Under Ankit's leadership, Infinitus supports 44% of the Fortune 50 and many of the largest healthcare organizations in the us. A serial entrepreneur, advisor, and investor. Ankit is a wealth experience including founding Ketra, which required by SimilarWeb, helping launch Google Play and co-founding Gradient Ventures, Google's AI Focus Venture Fund. Ok, welcome to the show.

Dan, thanks for having me. Excited for those conversations. Likewise. Before we dive into your scaly journey, can you give us the elevator pitch?

What does Infinitus do and who do you serve? We have built healthcare's agent communication platform to help solve the problem that I like to call a lack of communication technology that exists in healthcare, the five Ps of healthcare patients, providers, pharmacies, pharma, and payers. need to communicate with each other in order to get us on our healthcare journeys and do that in a as anxiety free way as possible. Unfortunately, a lot of that happens over phone calls and faxes and text messages and portals.

And we're building the AI agents to communicate all to connect these entities with each other. we serve some of the largest healthcare entities in the US from the biggest. National payers to the largest pharma companies that are trying to get patients access to medication at affordable, in affordable ways to some of the biggest health systems in the country as well. So in that thank you for that, sharing that context In that explanation, you outlined a couple of different modes, whether that's phone or faxes or, text.

Is, does Infinitus focus on one of those or all of those combined? Where is the business focus today? Maybe that's different from where you started. We started by focusing on the voice modality because that was the first problem that we wanted to solve.

As we have grown over the last six and a half years, we found that it's important to support our customers. patients, wherever they are. For some of them, the communication happens over phone calls For others, it happens through portals and chatbots. For others, it happens over SMS and text.

So we support a variety of different mechanisms today. So you founded Infinitus in 2019, and correct me if that's wrong, but which was before the pre-chat GPT era and Chet P made AI feel accessible to everyone as amazing as only three years ago. It feels like an eternity, but when you were pitching voice AI to healthcare organizations back then, what was the typical reaction that you'd get? An interesting story is we, when we were thinking about what to name the company.

we had the idea of calling it voice RPA. 'cause if you go back six and a half years, RPA was the buzz. It was robotic process automation. And you had companies like UiPath and Automation Anywhere that were the billboards around.

A number of the big metros and every buyer, every enterprise buyer in healthcare or beyond understood what that meant. It was being able to go to websites and an automatically entered information, extracted information. We said, Hey, we're gonna do this for voice. We're gonna automate processes that happen over voice, and we're gonna have voice.

RPA boy am I so glad we named the company Infinitus instead of voice RPA for two reasons, one. The underlying problem that we're solving isn't just in voice, it's a communication problem. And two, no one talks about RPA anymore. Right now, the equivalent of RPA or the next generation is called computer used agents.

But again, terminology keeps changing. What's important is to solve the the underlying problem. First, the first story is around the naming of the company and how we were thinking about the problem that we were solving. But the second thing was when we first started showing demos of what was possible, and this was in 2019, we were using large language models, which in retrospect don't seem as large as the ones today, but back then it was Bert and T five.

People thought we were doing black magic. They did not believe that a computer agent or a voice AI agent, as we now call them. have a conversation that was 30, 40, 50 minutes long. They were used to talking to Siri or Google or Alexa, which could barely answer one question, if any.

Well, Siri today still has that problem, by the way, but. yeah, right. And you know, but even the ones that had great experiences with some of the other assistant technology that was out there, didn't believe that an AI agent could talk for 35 minutes. And we were showing demos of.

These things I could talk for 35 minutes. And so there were some early believers who were eternally thankful for being the first to go live and then talk about it with others in the ecosystem. And then there were others who said, we don't believe this is real. And now many of them have converted.

And the chat GPT moment, in 2022. Changed the conversation. We, our sales process changed from not only selling the problem, you always wanna start by selling the problem that exists. But we had to, before December 22, had to prove that what we were saying was legitimate, prove that this kind of technology could exist.

We had to sell the problem and the technology. And I think after December, 2022, when chat GPT became part of the global zeitgeist. People believed that AI could do magic, and so we no longer had to sell magic. Well, I, going back to those early days, like, you hit some skeptic skepticism, I guess, I'm sure some of them were, in a mood to not be convinced.

But, maybe there were folks who were like, show me and I'll believe you. What would you have to do to prove to them in that era that it was real? That it wasn't just magic or you were just talking to somebody in the Philippines, tele operating this thing. Yeah, listen I think the reality is that we are lucky here in the Bay Area, and I was especially lucky.

To be here for the last 15 years when many of the early, what I think our AI companies built their technologies the best example of this in my opinion is Waymo. And they spent 12 years, 13 years before having a truly driverless car on a public road. They had safety drivers, they had all the checks and balances. And so when we designed our systems.

We did it in a very similar way. We were inspired by the Waymo's of the world, and when we talked to our customers, we said, listen, we have an AI agent. Here's a demo. We did live demos for our customers but we said every single call in the early days, we will have a human agent that is watching over.

To make sure it doesn't break guardrails or the human agent can take over if they need to. And so the idea of a tele operator was something that we built our systems with because that's what people had the comfort to buy. 'cause again, if you think about some of our customers in this pharmaceutical manufacturing space where there're, very and what they're allowed to say, what they're not allowed to say, their biggest fear, or the payers that the insurance companies in this country, they don't want AI agents that will hallucinate and hallucinations are a problem in 2025.

I can guarantee you that in 2019, before we had the kinds of LLMs we have today, it was a much bigger problem. Where we invested a lot of our effort was in the guard railing technology that we've built, but also in the human guardrails and the humans in loop that we had that got the ecosystem comfortable with us. Before they knew about, before they knew what all was possible. And so it was about, uh, building those assurances in sort of underneath that, Hey, like, we can still take over.

It's funny, I was listening to somebody and we have, uh, Waymo now here in Austin and it's a pretty magical experience, but, you know, you guys had it early out there, the Bay area, and so we didn't experience this, but you know, there used to be, you had to have a human operator in the driver's seat. They weren't doing anything, but they had to be there, right? So I could see a very similar story playing out of the B2B uh, world. That's funny.

Um, so then ChatGPT launches late 2022. I'm curious about that sort of just out a post, uh, ChatGPT era, like how did that change conversations with customers, prospects? You know, I mean, like I, you mentioned hallucinations I don't know that I had ever heard, you know, the term hallucinations. Applied to anything except a psychedelic experience before, uh, this or schizophrenic experience before, uh, the ChatGPT moment.

So like what did that enable, or like how did those conversations change for you given you'd already been in the game, like, and now everyone and their, mother is like, Hey, have you heard of this ChatGPT thing even made it into our political discourse? Yeah. I think it's been an interesting journey. You know, let's.

Start with some of the questions we started getting asked in meetings that we had with customers and prospective customers questions like, Hey, is this basically chat GPT? Can chat GPT do this for me? Like the build versus buy? Because now the IT teams go, well, can I just take chat GPT and do this?

Right. It's a fair question to ask. The second one, well, hey, I tried ChatGPT. When I ask it a question, it gives me different answers Every single time that I ask you the question, how do you make sure that.

Your technology does not do that because that wouldn't fit the guardrails of what the customers want. The third one. Now these conversations, those kind of questions started getting asked enterprise wide. So almost every large enterprise, prior to 2022, you had to go through the standard SaaS procurement process.

You have to get an MSA in place, you have to go through a security review, you have to go through financial diligence, and then you can go into sows and and start doing business. Well, you now started seeing. For the first time, ML review boards or AI review boards getting introduced into that sales process and often. These were groups of people that were centralized, that started having standard questionnaires that were being put together by five people that were the best at AI at these large enterprises.

That doesn't mean they were the best at ai, they were the best at AI at the large enterprises. And I can tell you at many of the large customers we have, we were the first company to go through an AI review board. And that was some of the most painful work our team ever had to do because we were being respectful of the process. But we also sometimes had to tell them that the way they were adjudicating what was appropriate or not was incorrect.

I'll give you an example. There was a large healthcare group whose review board asked us that were we able to, be bias free and how we were measuring bias. And so we had our standard ways of measuring them. Like, Nope, we need to know whether you don't have bias across.

And these were, let me specify. These were for calls that were being made on behalf of doctors' offices to insurance companies to see whether their patient's benefits covered the medication or not. Right. So you're calling a call center.

To get some information. And they were this group was asking us, can you see if, can you prove to us that the AI does not have bias across different ethnicity and across different age groups? And we said, well, no, because I have no idea what ethnicity the person at the call center is, and what their age is. And they said, well, this is our process, and if you can't answer this.

We can't your AI application through our AI review board. And we're like, well, do you want us to call, call centers tens of thousands of times and be like, Hey, before we get started, can you tell me how you how you identify yourself ethnically and what your age is? And they're like, well, that's what we would expect. And they're like.

ever does that. That's how you get people to hang up on you and say, this is a bad use of AI in our time. And so those were the kinds of battles we had to fight in the early days to get these AI review boards to realize that there's different buckets of AI applications and it's not a one size fit all adjudication criteria. Now they've gotten much better over time.

We have some core IP around something that we call a discreet action space because some of our customers require us to say things a very specific way. You cannot change a single word because their medical legal review requires it to them and any representative of those brands to to present themselves a very specific, well-defined way. Now, we all know large language models aren't gonna give you exact text in the way. It is meant to be.

And so some of the IP we have is around using large language models in order to know which pieces of text to identify and say out loud. And so there were a lot of things that we had to educate the market on. A Hmm. IP that had to be created.

So again the problems before the chat GPT moment and after the chat GPT moment we're different, but we're still in the early innings. Yeah, I, so that's a super interesting story. I guess, did you find have you fought that battle? I'm guessing there were better and worse ways or arguments that were more effective or less effective?

I guess what. What ultimately worked in that scenario? Was it just a matter of everyone sort of getting more attuned to how these sort of models work and so it's the problem just kind of, you know, fix itself or were there approaches when you start getting asked that question the third time or fourth time you hear it? Like that you were like, okay, like here's how we sort of address that, because I can imagine, you know, anybody shipping, sort of AI products is gonna run, potentially run into some version of this.

And it seems like it's an industry-wide issue. So just curious if there are any lessons you learned of that. So quite a few. One, I think like any sale.

important to have great champions who will pound the table and say, guys, this is solving a huge business problem. We need to not just be in analysis paralysis here and let's figure out how we get to a decision. And some of the early champions who helped us through some of these processes, what they did was they asked these AI review boards, what is the risk? That you are looking to mitigate and let's try to solve that risk.

So an example I gave the risk they were looking to mitigate was that they were serve different patients. In equitable ways. And our champions rightfully said, Hey, this is, we're gonna serve all the patients using this technology, but this is not a patient interaction. This is an interaction with someone in our call center.

We don't need to know whether this is, that first use case was only in English. And so we don't need to know bias testing across different languages because the call center agents are gonna talk to you in English. Now when we, as we've expanded and as we now do patient facing calls that are. Clinical in nature in multiple languages.

We collect the data around biases for people with different languages, different ethnicities, different age groups. But guess what? You have that information before you call the patient. Their for many of the large healthcare groups we work with, they've already collected ethnicity.

They already know the age because they have the date of birth. They already know the zip code, which can get you other social determinants of health. And so we can do that slicing and dicing. With the outcomes of the calls and provide them to AI review boards.

But it goes back to the three basic things I said. One, let's make sure that the criteria that are being used. specific to the use case, not just one size fit all the second you have a good champion who will really get to the heart of what the company is trying to mitigate from a risk perspective. 'cause at the end of the day, the goal of an AR review board is to reduce the risk that the company is taking on by bringing on a new vendor or taking on new technology.

And then the third one is really being open about. And thoughtful about the data that you collect, how you use it, and slicing and dicing it the right way. Hmm. of the things we've found is when you have that transparent communication with your champions and with your partners and your customers, everything just becomes much easier.

So, so, so good. Sales practice hygiene there. So identifying your champion and making sure that, you know, keep the, they keep the focus on the business problem that is being solved. Uh, really understanding the risk they're trying to mitigate.

And then doing your homework, like if you're gonna get asked this question on 50% of calls, right? You don't wanna be surprised every time someone asks it. It's like we should just sort of expect it and start collecting data and figure out a. A sufficient answer, if not a perfect one.

I don't think the industry in general has a perfect answer to this. I see reports probably at least once every couple of weeks now about, you know, different tests across the different LLM providers of different, uh, ways they, they bias their responses, et cetera. So the other, one of the other downsides you mentioned about the launch of chat. Was this a build versus buy?

How you have approached, uh, that discussion. I think there's a. A general sentiment that a lot of companies that maybe went down, oh, we could have our internal team kind of build a custom system with chat bt those or maybe fine tune sort of a model internally. Uh, there's some sense that maybe those projects didn't net the results that, uh, folks were hoping for.

I'm curious if there's either arguments or ways that you guys approached that problem. Yeah. I think a couple of things that I'll share here. First, I think building a demo has gotten easier with every iteration of this technology to the point where you can just describe what you want and some of the.

AI coding tools will build you the first demo of it. Hmm. I think it gives a false sense of security to teams when they can build the first version of the demo. I think what's gotten harder is to go from demo to production.

And so we spend a lot of our time all the things that go around a demo to take it to a scaled environment where we're fortunate that in the six and a half years of being. We've scaled our offerings to a scale that not many companies have. And so that social proof allows people to feel comfortable that this is gonna be something that they can get business outcomes and returns for. In a tight period that internal teams won't be able to.

So that's one. The second, I think as the underlying platforms become better and better, which you expect and want them to, like in every wave of technology, application or vertical SaaS companies, their value accrues. In the deep integrations and the deep integrations into the workflow, the security compliance, all in the specific solution suites for specific problems that are being solved rather than in the underlying technology under the foundational technology. And that motion is starting to happen even in in this space where any, again, anyone can build a demo.

the fact that you already have something that the customer can use right away is really powerful. Just like nobody today thinks twice about, which like once they need a database, they just click a button on your cloud provider and you have a database. The same stuff is gonna happen with your LLMs, and you'll be able to switch LLMs fine, tune LLMs in an easier way. When you think about things like text to speech.

can now give pronunciation dictionaries in a way you never could so that no one ever mispronounces infinitus or never mispronounces jein right. It's it's powerful that the technology has made those strides in the last few years. Yeah, there was a meme, I'm terminally online and there was a meme probably six, nine months ago how everyone was going to vibe code their own CRM or their own ERP and I, I. Just didn't bel I didn't believe it for a second.

I mean, I don't, may I, who knows, right? In three years we have a GI maybe every, it'll do everything for us and I'm just totally wrong. But, if you have, engineers or IT people, there's definitely folks in your direct comparative advantage that you could be working on that isn't building some other core piece of technology. That is another business is core sole focus.

So, but the AI infrastructure landscape is evolving. Fast. I mean, you've got now, agent companies you have platform companies like the LLM Foundation, model Providers, or Salesforce, and everyone's trying to kind of figure out where they fit. Like as now, right?

This has gone beyond, oh, chat BT is kinda this interesting toy to like. Yeah. Trillions of dollars of investment in this space and everyone trying to kind of sort out, you know, where the layers are, you know, is it at the chip layer, the data center, or the, you know, the AI enabled application or the foundation or the agents? Like how are you thinking about Infinitus positioning as this shakes out?

I think there's the system of record companies in healthcare. You can think about the EHRs. You can think about the CRMs like Salesforce or Veeva. the patient management systems that exist, we look at ourselves as a system of communication.

So it's one level above the system of record. So anything we do is based on data that comes out of the system of record and goes back into a system of record where the communication layer above it. Yeah. And I think it's a good point because I think there's this general sense that it kind of tied to the.

Previous thing I was referencing where system of records are gonna go away, but LLMs are not databases. They're terrible systems of record from an architectural level. So at some there will still be some system of record that's not, LLM, agent based. How is that positioning as the word these agent has got more popular folks are starting to talk about, hey, these are.

Potentially replacements for, personnel. Like, how are you seeing customers think about the value of AI agents, like in that frame? Is that how they're framing it to you? Are they thinking about it like, as part of their, like, Hey, we, instead of hiring, a hundred more call center reps, we could spend this money on, infinitus.

Like, how are you navigating that value conversation or hearing that from your customers? Yeah so a couple of things. I think healthcare is a very unique. in that there's a huge shortage of workers.

So there's a lot of work that isn't getting done. There's a lot of patients whose therapies are being delayed, and we know delaying therapy causes bad healthcare outcomes and therefore more cost to the to the entire ecosystem, which we all end up paying for in our increasing premiums year, over year, over year. And so the problem that we're looking to solve in healthcare is to augment. At the workforce that we can get patients on therapy as efficiently as possible so that we can make sure patients can afford the therapy and they stay a adherent to therapy.

So affordability, accessibility, and adherence, right? Those are the three pillars on which we really drive our communications platform that we have here at infinitus. So that's the first thing. The second thing.

There's a lot of work that is routine, tedious, repetitive, some of it administrative, some of it very clinical that our customers go. If we could have AI do this, the folks that we do have can do the stuff that it truly requires human empathy, that truly requires the ability to process things in a way that they're not having the time and space to do today. The third thing that I'll call out is there are certain processes. The first two I talked about were doing things that they've already identified that need to be done.

The magical stuff that happens is. Are processes that are being created net new processes that no one ever imagined in a world before ai because they were thinking from a position of scarcity, we don't have enough people. So they weren't creating more work for themselves 'cause they didn't have people to do it. So a good example of this is if a patient needs to get on chemotherapy or some kind of infusion, there's going to be a number of processes before.

The infusion, especially the first infusion to prep that patient. Do you have transportation? This is what to expect in your first infusion, and it is normal to feel nauseous after your first infusion for let's say 36 or 48 hours. If that continues, give us a call back, right?

Like those are the kinds of prep calls that the healthcare system has with patients today. And then we expect the patient, if something goes wrong or is terribly bad, that a patient who is anxious and in pain will pound the table, show up to emergency, make a phone call, and then we will react to it. Healthcare is designed to be reactive because of a scarcity of resources. AI flips that on its head.

We can now be proactive. We can now after an infusion, 24 hours after the infusion. Imagine a net new process where we call the patient and go, Hey. Patient you were.

You just had your first infusion. We know we talked to you and we told you that it was gonna be bad afterwards. How are you actually doing? Is there anything we can help you with?

And this patient might be at a eight out of 10 pain level and would not make a phone call 'cause it's not 10 outta 10 pounding the table yet. now might say, Hey, it's been pretty bad. I've not been able to keep any food down, or whatever it is. And the AI agent, because again, you have more resources than you, you could before in some contexts can give.

Have you taken your anti-nausea medication? Oh no. I totally forgot about that. Okay, well you should take it right, but someone or an AI agent checking in on a patient and changing that care journey so that it doesn't get to the 10 out of 10, you're proactive, you're ahead of it, and that is something that the healthcare system is starting to imagine, and Hmm.

powerful. I apologize I didn't ask earlier, but, you know, in several times this conversation, right? Healthcare, especially in the US is a complicated relationship between, uh, payers or insurance companies, clinicians, doctors and patients. Do you, uh, is Infinitus is your primary customer clinics and doctors, or is it also are you also selling to those other parties?

Yeah, so, I. I like to think about the healthcare system in the us. As the five Ps of healthcare, you've got patients first, and first and foremost you've got providers or the doctors and the health systems, et cetera, the clinics, pharmacies where you pick up your medications, et cetera. And some of these could be in person brick and mortar.

Some of these could be mail order pharmacies. The fourth are the payers or the insurance companies in the five fifth. The pharma companies or the manufacturers, or Hmm. the diagnostic companies?

We serve all of them in different ways, shapes and form. Because if you think about a system of communication, it's between all of these different entities. So we have made phone calls on behalf of pharmacies to patients on behalf of companies, to payers, on behalf of providers, to payers and patients. And so we're connecting those dots that are completely disjointed today.

But in terms of who Finis is sending invoices to it, it could be any number of those five at the end of the month. Correct. Not patients Not patients, not patients. but any of the enterprises.

of the others. And I'm spend all my time in the pricing world of one of the hottest topics in the world right now is pricing, AI and AI agents. So given you serve very different, four out of the five groups who may interpret value quite differently, is that. Implied different pricing models for your agents for each of those different parties?

Or have you tried to maintain a consistency or is it a mixed bag? You're sort of learning and iterating, which seems like what everyone's doing right now. Yeah, I know there's a little bit of learning and iteration but where we can, we try to optimize for outcomes based pricing, which is we want to deliver results. Now, sometimes that's really hard.

I. a for a buyer to get their head around because they're not used to outcomes-based budgeting. Today. They're used to human labor budgeting, which is hourly budgeting.

And when you say outcome, can you just give me an example just so we're talking the same language. yeah. So if if there's a customer who wants to make sure that a prior authorization status. Was completed or a pre or post infusion call was completed, you pay for the completion of that, Not just an attempt to do it.

Now that's very different than when you have humans doing work because you're paying for time regardless of outcome. And so in certain, in circumstances you want. Time-based pricing. In circum circumstances, you want outcome-based pricing.

And what we have found is, by and large in the back office, you want outcomes-based pricing. 'cause you want the system to be as efficient as possible, right? If there's a way to get data digitally over an API and get it back to somebody between a pharmacy and an insurance company. You want to optimize for that.

But if you're paying for time, well then even if I could get the data digitally, I'm gonna make a phone call or I'm gonna be incentivized to make a phone call because, and make that phone call as long as possible. Which isn't serving the ecosystem well, it isn't serving the patient well. 'cause it's delaying when they get the information that they need. Yeah.

right. And so in, in the back office, we find having an outcomes based model is better aligned to everyone's incentives. On the other hand, in the front office, when we think about patient facing phone calls, one of the problems you have today is because of scarcity of resources. You can hear the rush in the sound, in, in the voice of the nurse, because the nurse is already thinking about the next patient.

They need to call. The doctor is thinking about the other calls they have to do before they can wrap up for the day. in the case of a patient facing phone call, you actually wanna charge by time because it's okay. the patient is given an extra five or 10 or 15 minutes, if they actually have those questions, because you are treating time of a nurse or a doctor for Mm of an AI agent, and you're giving the patient access to resources in a time of anxiety that is gonna serve them better in their healthcare journey.

And correct me if I, uh, am misunderstanding, but, so within that lens, you painted within, say a doctor's office or clinic. That front office, back office, they could potentially have a preference for one pricing model in, in the front of the clinic at a different, in the back of the clinic or depending on the type of call that they're making. What have you learned about like, straddling that? Is your approach been like, well, we're gonna kind of give, we'll just have sort of those two as, as different defined sort of agents with their individual pricing model?

Or have you tried to standardize yeah. I think we started from an outcome-based perspective because I'm a big believer that if you can drive the right outcomes, everyone's incentives are fully aligned. Yeah. is, a, you charge for outcomes, the per unit price might be higher than the time-based pricing.

Because you don't succeed a hundred percent of the time. And sometimes for the buyer who doesn't know what, for the first time buyer who doesn't know the expected outcome rate, that might be hard. That might be a hard pill to swallow. Even though the end result might be better than what they currently have.

And so that's a discussion we have. And in some cases you go on a time-based model for some period of time to benchmark everything, and then you move to an outcomes-based model. At the end of the day, you want your human resources to work on things that humans should work on, and you want AI to work where AI should work. Yeah, I think that's a really good point that you brought up.

I'm not sure if or follow Dave Kellogg. He is amazing. He gave a presentation recently talking about outcome based pricing and he used the example of with his dating apps like. Tinder, for example.

And so if you say like, okay, my, my outcome is pure, it's what it is now. Either monthly or yearly or I think they have even a weekly option, right? That's one price. If you price per date, you know, you have to look at like how many matches convert to a date or how many matches you get per month, right?

To convert that to number of dates if you wanted marriage, it ends up being like, you know, whatever point. 5% of customers hit that outcome. So, you know, at the end of all that math, if you wanted to price your dating app based upon like a successful marriage, you'd end up having to charge the person like $15,000. Right?

And so, I don't know many people would be willing to sign up, but if it was a really great talk, canned example of yeah, people think about outcome is really. You know, great. But then when you realize, you know, it could take many attempts and tries to sort of reach that outcome, you could end up with sticker shock at the end if you're not, if you're not careful. So I imagine that's a little bit of what you were sort of, navigating in some of those conversations.

The one thing that I'll say is different than dating is I think for. most disease states and most therapies where you see the most need for these kind of technologies, they're chronic in nature. hmm. so it isn't something that's going away in the interim.

And I think that dating is also an interesting industry to study Oh, it's very interesting. well, if the underlying. Platform is successful. They churn customers.

Right. And so it's a Yes, acquisition story versus, you I. healthcare enterprise is, they always have, as long as there's humans, there's going to be people to take care of. And so, the dynamics are slightly different.

Yeah. Yeah. You can actually have it perverse incentives in the dating market. You get perverse incentives in any market, right.

We subject people to more potentially tests or, uh, things that they need, right? Because we get to bill, medicare for those things. Right. Um, you know, not saying that's done, uh, regularly, but, you know, have heard of those things before.

I do wanna pivot. I, I wanna focus, uh, you've navigated this incredible, uh, market evolution, being the only game in town, you know, mid of 2019, you know, almost, uh, competing in crowded AI landscape, but beyond sort of these external market dynamics, you've also built the organization in a pretty unconventional way. And so I wanna talk about this. So like you've.

Maintain. From what I understand, again, correct me if I'm wrong, you maintained a titleless organization for almost six years now, so what led you to try this approach in the first place? No. One of my mentors Google.

Roger I've learned a lot from him about building company and culture over the years. And he had this blog post 2017 which said which was called Titles. And I curiously read it. And one of the areas it talked about is how titles caused more problems than they than they solve.

Because once you create titles. starts becoming motivated by titles. If you are an engineer two, your goal is to get to engineer three. And once you have titles, it leads to entitlement because people start trying to assert power based on title and make decisions based on title.

And I still remember maybe a decade ago maybe longer than that. Jack Dorsey, who was then the CEO of Twitter a note to all employees of Twitter saying be in a meeting and make a decision where I'm not in the meeting. And someone says, well, Jack wants it that way because if I wanted a certain way, I will come and talk to you. But I don't want people using my name in a way to make decisions or push a certain viewpoint.

And at the heart of that is the fact that decisions shouldn't be made because of a title or a position. They should be made because they're the right well, debated decision and defined decision for a company. One of our company values is disagree and commit. It's a leadership principle at Amazon.

We take it a step further. We say, if you disagree and commit, and then you have to become an evangelist for that for that decision. Title, the problems with titles continue, right? Once you get someone on a title train, if you will, it's always trying to figure out what do I need to do to get the next title, even if it's not in the best interest of the company.

Because then they go, Hey, we need to we need to optimize. For me going from senior manager to director, director, the senior director, senior director to vp, and then you start having rubrics of, well, in order to get to this, you need to manage so many people. And VPs at our company typically manage. people or like you come up with arbitrary rules of that sort.

And I think it, it pushes the wrong incentives and I think it all comes down to incentive structure. And also, you know, probably the last one that I'll talk about and there's many many more and go calls block posts on this, it's great, is it's really hard to UNT title. And as a growing startup, when you give someone a title, it's really hard to take it away 'cause it's a hit to the ego. If you are made a manager or a director.

And you were told, Hey, now we've scaled to a level where you shouldn't be a director. You need to go being a senior, go back to being a senior individual contributor, you're probably likely to lose that person because they have to now face their peers. In what seems like a potential demotion. And we've seen this as we've grown, from, a couple of people to now just over 200 people.

Having a titleless culture has helped us in tremendous ways. We've had people go from IC roles to management roles and back to IC roles. And we do this because it's titleless and no one ever thinks it's a hit to their ego. So I totally get the incentives driving.

Unhelpful behaviors especially if, I think there's a, you are at Google, so maybe you could speak to this, but maybe it's apocryphal. But, uh, there's some story that like, the v ones of Google products are because they're going to be, 'cause there's something on the promotion chart that says you have to go to the next thing and ship new products is to get promoted. So they never really improve their existing products that much. We'll leave that aside.

Uh, but the so I think, you know, there's, uh, many examples we could point to. I don't mean to pick on any one company there, but. Let me play devil's advocate. So like, how does this actually work in practice though?

Like, how do you go, how do decisions get made? So like, I used to be in the product organization, right? And so, right. The chief product officer is like, trying to maintain some coherence of, strategy and there's a bunch of different things, happening.

And so, it's a title. It could also be handy 'cause it lets everyone else know. Like, I've gotta make sure that this is aligned with this person's view so that we're not off, in our own lane doing something completely wacky. So how does, like, how do you resolve that type of situation?

That's a great question. So a couple of things. There's actually two problems that you have to solve, one and the second. Is career growth and compensation growth which sometimes are tied to titles.

And so I spent a lot of time kind of thinking through this and designing parts of our our system for, to make sure we solve for those two. So a titleless culture doesn't mean that you cannot have leaders. So when, if you look at my LinkedIn, it says I'm company lead 'cause I lead the company. For someone who leads the technology team, they're.

Or engineering team, their title is Engineering Leader. If you lead like we have a patient facing AI agent, so there's a patient AI agent, product lead, Hmm. right now that can change now, that person in for the next quarter is leading this effort. Now, if they go and they lead something else, if you're leading a project or you're leading a group of people for that period of time, you have the.

You are known as the lead for that effort, but you're not known as a director or a senior director or VP or SVP, so it's very clear who the decision maker is. For certain kinds of for decisions that are needed to happen. And then, the product leader like in the example you gave overseas, the AI agent products, the under which you've got patient facing, payer facing, provider facing agents, or SMS agents versus voice agents. Each of these different projects can have a lead or, there's equivalence on the engineering side or on the recruiting side.

And so you always know who the decision maker is and how those leaders roll up into the next level of leaders. But it's not a it's a decision making hierarchy rather than a a seniority hierarchy. So you could have someone with way more experience in their career who is responsible for one. Kind of decision versus what seems like a peer with different level of experience, but is the right decision maker for that pod, if you will.

Got it. Well, you said there were two things to note. So one was decision making and then the other one. was compensation, right?

If, for any new hire, they go, well what, what do promotions mean? And what, like, how do I make more money? Like it's a real question right over time. And so internally and.

Privately between a person's manager and them. There is a leveling system where you have goals, where you have what's expected of you and what you need to do to get to the next level. And that is not a public leveling. It's private between a manager and their report.

So that way you can grow, you can have goals, and as you make progress there you can have compensation changes as well. When we celebrate promotions internally, and we send out a note in our internal chat spaces and it says, person X was promoted from member of technical staff to member of technical staff in the last year. They accomplished A, B, C, D, and E. Please join us in celebrating person X.

Well, I'd be interested to check in with you in a couple of years and see if you're still running the system. I definitely believe, you know, title inflation is a real problem. I mean, I've worked at companies where, you know, we had like VP and then SVP and then EVP, and then the, there was much SVPs, well, I get promoted, so they created like a group VP thing and it was just like their jo day to day didn't change. It was just to give that person a bump.

Uh, and I was like. This is like, a weird game of charades that we're all involved in. So I don't know if your approach is the solution, but it's an approach and I'm rooting for you to succeed. Well, I appreciate it.

I'm also very curious how it plays out over time. The other thing that I tell my team is, listen, whether you are a director, senior manager, or VP at a company at our size doesn't really matter. What matters is if we succeed or not. 'cause if we succeed, people will say, oh, you worked at Infinitus, that incredible company.

We want to hire you regardless of what you did. And if we fail, your title here won't matter either. And so I think, we're rooting for our success. And at that point, hopefully titles aren't a problem either.

I love that. Yeah very true. I think some people wanna probably go scrape their LinkedIn history, even if they were a C-Suite title at companies that maybe didn't do so well even after they left. I get this has been awesome.

I would love to talk to you all day, but in the interest of time, we're gonna start rapid things up. I want to pivot to a couple of rapid fire close out questions. Is that okay? Absolutely.

Alright. What's your favorite business book or podcast right now? I've been enjoying the depth of interviews in, acquired a ton the last few months. Acquired is a podcast, correct?

That's the podcast. Yes. Yeah. Awesome.

When you think about all the spectacular people that you've had a chance to work with as anyone who just pops to mind, that's had a disproportionate effect on the way that you think about building growing companies, being a. My wife I've what it takes to support someone who's doing their own company, and I think it's it's. As much a part of success or a journey as anybody. There's a lot of things I learn from many of my mentors, but I think it's often my wife that grounds me and makes me think about how to think about the process and the journey with other people rather than just dollars and cents.

It's excellent to have that grounding family, very important because the business world could be quite brutal and so it's good to have a person to go home to. How do you continue learning and growing as CEO? Is it like specific resources besides acquired or other peers, mentors that you. Yeah, You work with or learn from?

I spend a lot of time in peer groups learning from other CEOs or folks that have been through this journey before. But the one that I, I. learned the most from is actually doing angel investing because I get to see companies go through their journey over time. Some that go faster than we are doing here at Infiniti.

Some that are going slower and. In the last five years, I've probably done 80 investments and through those 80 CEOs and some that have changed over time, I've learned a lot well, I hope your portfolio is doing well. If our listeners want to connect with you, learn more about Infinitus how can they do that? Well, all the traditional social channels on LinkedIn Infinitus systems or myself at Anki, Jane.

Awesome. I'll put those links in the show notes for our listeners. Everyone that wraps up this episode of Sask Felix here. Thank you to Anka for sharing his journey and insights.

For our listeners who found Ancos Insights valuable, please leave review and share this episode with your network. It really helps the podcast grow.