Ankit Jain, Infinitus, on applying AI technology to unlock time and efficiency in healthcare
In this episode, I sat down with Ankit Jain, Founder and CEO of Infinitus Systems. Founded in 2019, the company automates tedious and time-consuming administrative processes across the healthcare ecosystem with AI to free up clinician time and improve access, adherence, and affordability. Infinitus is the only solution that automates the entire phone call process to complete benefit verifications. Infinitus returns that hard-to-find information so providers have the most comprehensive view of treatment coverage and can confidently move forward with a patient treatment plan. The company has raised money from Google Ventures, Kleiner Perkins, Coatue, and a number of other healthcare luminaries.
Ankit and I discussed:
- The company’s “spark story,” and how a Google digital assistant for restaurant, spa and salon reservations was pivoted to fill an unmet market need in healthcare
- The importance of freeing up time spent on administrative workflows for clinicians to serve patients
- Applications of AI in healthcare, both on the back-end and patient-facing/clinical sides, and how this will evolve to continue solving for areas of unmet need
Beginning— 4:45: Ankit’s Background and Spark Story
Ankit started at Google in 2010 leading Search & Discovery features for Google Play, after which he founded several companies of his own in the tech space. In 2017, Ankit returned to Google as Director of Engineering and helped launch its AI-focused Venture Fund, Gradient Ventures, through which he invested in vertical and horizontal applications of AI, around the time the Transformer paper was written about neural network architecture for language understanding (content which later fueled the “gold rush” of AI including ChatGPT, Bard, etc.). As Ankit increasingly focused on investing in healthcare around 2018, Google demoed Google Duplex at its annual developer conference, which made use of a voice assistant to automatically make online spa, salon, and restaurant reservations. With this tool, leveraging speech recognition, natural language processing, speech synthesis and AI in a tightly-coupled loop, Google Assistant could make phone calls when needed to reach a human on the other side to complete reservations, and then confirm the reservation was made.
Ankit was very impressed by this innovation, and took it home to his wife, Shailvi, that evening. Shailvi had spent the last decade and a half working in healthcare, focused on health ecosystem administration, including serving roles at a home health and hospice clinic, at the oncology department at Stanford Health, at an urgent care center, and at a large medical device and pharmaceutical manufacturer. Equally impressed by the technology, Shailvi expressed that she wished this same technology could be applied in healthcare — across every single institution she had been a part of, people sat in the back-office making phone calls to exchange data and “move the chains of healthcare forward.” According to Ankit, his “aha!” moment occurred after hearing Shailvi express: “I wish someone would automate all of this, because no one joined healthcare to wait on hold, to collect information over a phone call….we joined healthcare to serve patients. If we could automate all these phone calls, we could unlock time for all these [providers] to spend time with patients and their families.”
4:46–9:40: Unmet Need, Illustrated through a Patient Story
Further highlighting the unmet need in the industry, Ankit shared a story: “Let’s say I’m a 65 year old male who’s had diabetes for most of my life, and I’m starting to have some difficulty with my vision. I go to an ophthalmologist and they do an exam on me. And they tell me that I’ve got something called macular degeneration.”
Patient: “That’s a big set of words. What does macular degeneration mean?”
Physician: “Don’t worry, there are medications for macular degeneration. These medications are specialty medications, they can be very expensive. They can range from $10,000 to $100,000. And to get them administered, you come back into the office every two to six months, and we inject the medication into your eyeball.”
Patient: “Wait, wait, wait, you, you do what? You inject the medication into your eyeball? That sounds scary….”
Physician: “Don’t worry, this has worked for most patients, and it’s something that will really make you feel better. I really believe based on the diagnosis, this is the right medication for you.”
Patient: “Great, how do I get started?”
Physician: ‘Well, it’s not that simple. Because this drug is so expensive, I can’t just inject a $10,000 medication into your eye, and as a doctor, I have to make sure that if a vial of medication is purchased, I will be reimbursed for it. Either your insurance will pay for it, or you will pay for it, but someone has to pay for it. So before we can schedule your first appointment, my back office staff is going to make sure that your insurance, if you have it, covers it. If they do, we will then see what percentage or amount of it they will cover, therefore I know how much our office has to bill you. If you can’t afford it, we will talk to you about many financial assistance programs and help you get the financial aid you need. Sometimes, your insurance company will ask us to prove that this is the right medication for you, and that we’ve tried all the other simpler, cheaper options already — this is called a prior authorization, so we must submit all the paperwork and ensure it gets approved. Again, the goal here is to make sure that you get the right medication in an appropriate amount of time…but we are also paid for our services, and the product that we use on you.”
Patient: “Okay… great. So can we get that started? And can I come back tomorrow?”
Physician: “Well, no, no, no, not that simple. We can see you in a week or two, and we will call you to schedule your appointment, as we must give our team time.”
Clearly, as the patient coming to terms with a big, scary, new diagnosis you’ve never heard of before, and processing this idea that someone’s going to be injecting you in your eyeball, all while taking in the fact that you might need thousands of dollars of medication that you may not be able to afford… oh, and the fact that you could be feeling better today, but that’s not going to be the case, because you might have to wait a couple of weeks for the processes to unfold on the back end as providers and administrators work through all the paperwork… there is a lot of room to streamline this process.
UNMET NEED: Infinitus dug into this issue as one of many issues within the patient journey. Ankit and his team realized that, while much of healthcare has been digitized over the last 15 years, from electronic health record (EHR) systems to patient portals and more, many of the digital connections required to transfer information between parties still occurs via faxes and phone calls, which are very inefficient. This is where the Infinitus system comes into play.
9:40–19:25: Infinitus Solution
- Overview: Infinitus’ digital assistant can make phone calls automatically on behalf of a provider’s office that must figure out the benefits involved in a specific diagnosis or therapy, enabling these providers to converse seamlessly, collect required information, and move the patient forward efficiently and swiftly in their healthcare journey. Rather than have a physician, nurse, or administrative assistant make a phone call, Infinitus has built API connections between the patient’s EHR or relevant provider CRMs and the phone call system, sending the request through Infinitus, which makes the call automatically on their behalf. The Infinitus solution has been used thousands of times a day on behalf of tens of thousands of providers around the country, streamlining calls that on average take 37 minutes, and may last a couple of hours.
We had the pleasure of listening to a real call made between the Infinitus digital assistant, Eva, and a large insurance company in the Midwest, BCBS, on behalf of an ophthalmologist prescribing a medication for macular degeneration — highly recommend listening to the episode to hear the impressive tool in action!
“Imagine being a person who joined healthcare to serve patients. And instead, you’re in the back office making these calls all day, every day. Maddening. This is why our patient wait times are so long.”
- Establishing a Training Dataset: At the company’s inception around 4.5 years ago, the team looked for many datasets including providers performing these kinds of phone calls. They got access to about half a million phone calls conducted by humans, mostly for benefit verifications and prior authorizations. As the team sifted through them, they realized every single call was done slightly differently, accounting for natural differences in human interaction, as well as errors. After learning a lot from these, the team decided that the best way to automate such a complicated process was to step back and design a more standardized system, building standardized modules to drive the automation and conversation with the party on the other side (usually insurance).
- Partnering with Payers: As Ankit explained, “we’re not in this to just automate more and more calls and have machines talk to humans on the insurance side…we want to be partnering with [payers] as well, so that if they can provide us data digitally, we’ll eliminate the phone call.” The value proposition here for payers is to both 1) unlock time on the provider side that is better spent serving patients, thereby improving patient outcomes, and 2) free up time on the payer side as they feed Infinitus the data digitally, making it so that provider advocates don’t need to be on the call to exchange data. Ideally, call times will get even shorter as payers continue to provide data digitally.
It all comes back to our mission, which is to create more time for health care. And so, our success is measured by how much time we’re unlocking, for providers, for pharmacies, for payers, for everybody in the healthcare ecosystem, and reducing the time to therapy for patients. Some of these things you can measure in the short term, some things require more lengthy time studies to figure out impact. But on the payer side, the ROI is actually pretty straightforward — it’s a pretty immediate ROI for a payer not to have to pay for a call that could have been 30% faster than the equivalent human call.”
- Condition-Specificity: Since the Infinitus system interacts with many different customer groups across thousands of therapies and over a dozen therapeutic areas, it must “learn the idiosyncrasies that exist in different niches of healthcare, because things are done differently across oncology, rheumatology and neurology.” Benefit structures differ depending on whether a patient must be referred to an oral medication, an infusion, or other forms of therapy, and each of these may be either in or out of network depending on the specific patient. The Infinitus team mapped out all these different pieces by condition to understand how they work, embedding them into their dataset.
- Enabling Efficiency: After accounting for condition-specifics, the team acknowledged a need to understand that each person receiving a call would have a different style, but that ultimately, their job was to answer questions and to move onto the next case — “So the more efficient you are, the more you’re helping them succeed.” The team has done side by side studies of humans vs. Infinitus digital assistants making these calls, and found that Infinitus saved 30–35% of time on average, driving physician happiness and enabling them to hit their metrics.
- Triaging Workflow Issues: Infinitus has built a very robust human and loop infrastructure such that if Eva, the digital assistant, ever needs help, it can request help from a human representative. Ankit explained, “You can think about it analogously to the self-driving cars that are running around many cities in the country, which sometimes have a safety driver in the car to help the machine when it goes, ‘I don’t know what to do here. Can you please take over?’” Every day, the number of interventions that must be made by the Infinitus team improves as the algorithm and system learns.
19:26–23:42: AI Across the Healthcare Ecosystem
Figuring fewer leaders would be better equipped to speak to the rapidly evolving AI landscape in healthcare than Ankit based on his experiences at Google, I asked him to share some of his expertise. Ankit elaborated on several of the use cases for AI in healthcare:
Perception: AI allows us to understand data, images, text, and apply this all to medicine. A few examples:
- In radiology, AI and computer vision can elucidate which few scans out of thousands of scans a radiologist should look at in order to make a determination of what happens to a patient; or, even better, automatically determine how to triage the next step of a patient’s care
- Large transformer models can enter medical charts and build a summary of a patient’s history such that, when a provider walks into an office to see a patient, they have consolidated view of that patients’ past encounters
- Digital scribes can listen to a conversation between a patient and doctor and take notes in a patient’s health record, and sometimes even take it a step further and code the encounter as well for billing purposes
Creation: AI, similar to humans, can also perceive and learn from the world and then apply this towards creating new content, as in the case of generative AI
- Several life science companies are coming up with new drugs, creating molecules and testing them out using generative AI. One such company is creating digital doubles for clinical trials, applying these digital doubles to placebo and control groups to see whether the medication has a positive impact, and removing the need for patients to participate in the trials
- Prior authorizations can also apply generative AI, leveraging machines to sift through documentation, look for “needles in haystacks,” consolidate relevant information, and present it to a nurse who makes a final adjudication.
- While back-office operations have clear use case for generative AI, several front-office operations also benefit from this technology both pre- and post-encounter with a patient, whether replacing the need for a patient to fill out forms on a clipboard at the beginning of the appointment, using a digital assistant to make a follow-up appointment, or pulling up old patient information at the time of a visit during the encounter.
Ankit also highlighted several challenges with AI in healthcare, including the fact that piloting new technologies with new vendors can be challenging, because there are privacy concerns associated with revealing vast quantities of data. With generative AI, it is possible to create a whole set of new data that can be used in a synthetic or demo environment, and leveraged to conduct a pilot that tests out the efficacy of a solution before patient data is incorporated.
Overall, applied correctly and safely, Ankit feels strongly that we can use AI “to improve clinical processes, drug development, and administrative processes…on the provider side, on the pharma side, [and] on the payer side.”
23:43–25:35: Consolidation in a Saturated Healthcare Market
I asked Ankit how, “In an era where so many startups and point solutions have emerged to tackle issues in healthcare, and as technology changes so rapidly, what are some signals that we can look towards to indicate whether a particular solution will last or win in the market?”
Ankit agreed, stressing that there are 8,500 digital health startups, each of which is trying to rapidly acquire customers and health plan logos, and expressing that he expects to see market consolidation “as folks understand how not only to develop the technology, but then deploy the technology, and gain competitive advantage.”
Additionally, Ankit highlighted the challenge of vendor burnout: “From the buyer’s perspective, every additional vendor costs customers a million or so dollars just to manage…that’s a huge cost to them. In a world where you’re trying to trim down costs, the best thing you can do is go to an existing vendor and say, ‘hey, can you also do X?’” This further lends credence to the idea that consolidation will continue to happen over the coming years. “If you’re on the vendor side, I think it’s important to figure out whether you are the ‘Puzzle Maker’ or a ‘piece of the puzzle,’ and really hone in on what your mission as a founder is.” Accordingly, “as a founder of a company, figure out the impact you’re hoping to have, and how you’re going to continue to grow that mission from where it is today to where it could continue to go in the future.”
25:36–28:47: Keeping Pace with Change as a Healthcare Business Leader
As AI continues to evolve rapidly, business leaders must continue to innovate in order to keep pace with this technological change, and this is no different in healthcare.
“There are multiple pieces to building a technology business. I think, often, because of the media, people think of an AI solution as the AI model itself. [But] the model is [just] one part of the solution that the customer is buying… they’re [also] buying a deep integration to the workflow, they’re buying the understanding of the underlying data assets, they’re buying the support and services that are being provided to them. And if you design the technology solution well, there’s an abstraction layer between the rest of the infrastructure and the model itself. And I think it behooves any technology company to constantly be trying out new models and seeing what performs better, and how the model models are either continuing to deliver the highest value, or drifting in a negative direction, [all while] keeping the end customers value proposition in mind.”
When it comes to innovation, Ankit’s role at Google and in the broader technology industry certainly shaped his role as an entrepreneur, particularly as it relates to understanding the value of a brilliant team. A strong team, he feels, is what differentiates many Silicon Valley startups from most other enterprises out there. However, healthcare-specific skill sets are also an important part of building a strong team:
“You have brilliant minds that are looking to have an outsized impact on the industries [in which they] get involved. Most of our product and engineering team did not come from healthcare, but they are very passionate about having an impact in the healthcare ecosystem. The flip side to that is, healthcare is unlike a consumer company where you can’t just build, launch and scale something in a couple of weeks or months. The sales cycles are long, the decision makers conduct detailed ROI analyses, [and also require] deep relationships, which just takes time…As one of those companies [with] outside technologists coming in to change healthcare, it’s about thinking in conjunction with those that are healthcare insiders, and saying, ‘how can we improve this?’ Because yes, we’re all passionate about doing it…but let’s do so in a way that healthcare can actually adopt and scale.”
28:48–30:59: Building Towards the Future
With respect to unmet need in the healthcare industry, Ankit feels that too much time is still spent on many processes, whether clinical, administrative, or research, for people to document information, consolidate information from disparate systems, or share information across the ecosystem. “Building those connective tissues across the many different parts of healthcare, so that information can flow freely, is something that’s desperately needed across healthcare.”
Accordingly, as Ankit thinks about forward-looking innovation, he feels that “the power of modern AI technologies is to serve as a very efficient translation layer. [This is] where I see the most upside in the next few years to come. How do you reduce friction? How do you reduce burden so that those in healthcare can really do what they originally came there to do, which is to deliver health care?”
As Infinitus maps out its path towards playing a role in this journey into the future, Ankit harkens back to the company’s original vision, which is to “create time for healthcare.” He clarifies, “it’s not just meant to be time on phone calls — it’s time in general.” This speaks to Infinitus’ potential to extend beyond making phone calls, and towards enabling other time-constraining processes across the patient journey so that “we can make healthcare about healthcare again.” After all, we’d all benefit from freeing up clinicians to do what they are actually meant to do — delivering quality care to patients in an accessible manner.
31:00 — End: Closing Entrepreneurial Advice
In closing, Ankit shared one of the best pieces of advice he received from one of his board members two years ago:
“Accept that, if you want to accomplish your vision, it’s going to take a long time. I think it’s very easy to go for a quick ‘burst of flame.’ But sometimes that flame grows very large, and sometimes it flames out. And in health care, a big part of having a long-term impact is just being there for the long term, and committing yourself to your mission, to your vision. Making sure that you can build a business that is self-sustainable is very important.
This explains why, when thinking about Infinitus as a business, Ankit and his team constantly consider the question, “how do we make this a business that isn’t dependent on outside capital, [such that] it can become profitable and actually scale in a sustainable way?’
We are very appreciative to Ankit for sharing his wisdom with us on this episode of The Pulse Podcast! Subscribe for our new releases on Twitter, Spotify or Apple podcasts.
