How Infinitus Built Healthcare Voice AI Before ChatGPT: Ankit Jain on $102.9M Raised and 30-Minute Phone Calls

September 21, 2026

TL;DR: Infinitus Systems builds AI voice agents that make the phone calls healthcare runs on — benefit verifications, prior authorizations, co-pay foundation checks and, increasingly, calls to patients themselves. It was founded in February 2019 by Ankit Jain and Shyam Rajagopalan, friends since middle school and co-founders of a prior startup, after Jain's wife watched Google's Duplex demo and told him the technology was being wasted on restaurant reservations. Infinitus made its first production calls in late 2019, years before ChatGPT made voice agents fashionable. It has raised $102.9 million, most recently a $51.5 million Series C led by Andreessen Horowitz announced in November 2024, with Kleiner Perkins, Coatue, GV and Memorial Hermann Health System also on the cap table. Average customer contracts are over seven figures. Jain told the story on The Product Market Fit Show.

What does Infinitus do?

In the US, a doctor cannot simply prescribe an expensive drug and move on. Someone has to call the insurance company to find out whether the drug is covered, what has to be proven first, and what share the patient pays. That is a benefits investigation, and it is followed by a prior authorization and then repeated calls to check its status.

Large provider groups, pharmacies and supply chain companies employ hundreds or thousands of people to make these calls. Infinitus replaces them with AI voice agents.

Key stat: The average call Infinitus automates runs over 30 minutes and follows detailed standard operating procedures. As Jain described on the show, roughly 25% of the time the insurance agent gives incorrect information — so the agent has to check answers in real time against what it already knows and ask the human to confirm.

The product has since expanded in three directions: to pharmacies and co-pay assistance foundations, to patient-facing calls preparing people for a first infusion, and to the insurers themselves, several of whom now hand over 50 to 70% of the data through APIs instead of reading it aloud.

Who founded Infinitus?

Ankit Jain is co-founder and CEO. Before Infinitus he helped start Gradient Ventures, Google's AI venture fund, launched shortly after the "Attention Is All You Need" paper — a job that gave him a front-row view of what the technology could do. He had also founded and sold a company before that.

The idea came from his wife, who has spent her career in healthcare, while he was showing her the 2018 Google I/O Duplex demo.

"Ankit, incredible technology. Good demo, but why do you need this to make spa reservations? I wish someone would bring this to healthcare." — Ankit Jain, Infinitus, recounting his wife's reaction

"And she said, great technology. What a terrible use case." — Ankit Jain, Infinitus

He validated it with a conflict of interest he later had to explain. Still at Google's fund, he cold-called prospective customers saying he was diligencing a company that automated back-office phone calls.

"The first one said, we have 3000 people doing it. The second one said 400 people. The third one said a thousand people." — Ankit Jain, Infinitus

Every one of them asked to be introduced to the company. A few months later, Jain went back and told them it was his.

Key stat: Five months elapsed between the conversation with his wife and the founding of Infinitus in February 2019. First production calls followed roughly six months after that.

His co-founder Shyam Rajagopalan built the first prototype as a Wizard of Oz test: a fully robotic voice, with the founders typing the responses, calling a major insurer to ask about a patient named Bruce Willis.

"And then they were like, let me look that up. And we're like, looking at each other. Oh my God. Like they're actually talking to us about it." — Ankit Jain, Infinitus

The question was never whether the agent sounded human. It obviously did not. The question was whether anyone would stay on the line.

How much has Infinitus raised?

Verified as of September 2026:

Seed: just under $4.5 million, 2019, led by Kleiner Perkins with angels, as Jain described on the show.
Series C: $51.5 million, announced November 2024, led by Andreessen Horowitz, with Memorial Hermann Health System joining and existing investors Kleiner Perkins, Coatue and GV participating.
Total raised: $102.9 million.

Key stat: Infinitus's average customer contract is over seven figures, and its first customer grew 10x over four years.

How Infinitus found product market fit before voice AI worked

In 2019 there was no off-the-shelf stack, so Infinitus built most of it: speech to text, the language understanding in the middle, and voice synthesis, plus original research on combining audio and text models, because dropping voice to text throws away the signal that tells you a person is annoyed or confused.

"In 2019, 2020 people thought we were doing black magic." — Ankit Jain, Infinitus

The deployment model came from self-driving cars.

"We took a lot of inspiration from our friends over at Waymo. I think one of the things Waymo did very well in the self-driving world was they said, let's get a car on the streets as early as possible. And if you need to have a safety driver, that's okay." — Ankit Jain, Infinitus

Infinitus shipped an agent that could raise its virtual hand and ask a human for help, then used every one of those interventions as training data. The guardrails built for 2019-era models — when hallucination rates were far worse — are the same guardrails that now give compliance teams confidence.

The first customer was a Fortune 10 healthcare supply chain company with thousands of people doing these calls, one of the firms Jain had cold-emailed during his diligence. Getting a company that size to pass personal health information to an unknown startup with no other customers took a full security and compliance program and a long, hands-on rollout — starting with a small share of calls and side-by-side comparisons before scaling up. Jain said the customer has publicly described the results, including roughly 10% better accuracy than its own teams.

Key stat: Jain dates the first real product market fit signal to about one year in, when the first customer was live and working, which gave the team the proof to sell everyone else.

Never miss a founder's PMF story

Subscribe to The PMF Show

Why insurance agents did not hang up

Early on, roughly 30 to 40% of calls were hung up on. An early-Uber friend reframed it for him.

"You should never think about the 30 or 40% that get hung up on. You should think about the fact that 60% of the time you've moved into the future." — Ankit Jain, Infinitus

The acceptance rate came down to context. To reach an insurance agent at all, you have to authenticate in the phone tree, wait on hold for half an hour, and ask relevant clinical questions. Spammers do not do that.

For three years, the callback number Infinitus gave out on every call was Jain's personal cell phone. That produced the company's most useful sales channel.

"Hey, I am a supervisor of a call center team here. Who the hell are you and why are your robots calling my team?" — Ankit Jain, Infinitus, recounting a call he received

Jain would explain what the company did, then ask what the call center agent had thought of the experience.

"He is like, I hate to say this, but they kind of liked it." — Ankit Jain, Infinitus

The agent was polite, efficient, easy to understand, and never yelled. Call centers also came to recognize Infinitus by its consistent phone number and single voice. Several of the largest insurers in the country are now customers, having first heard of Infinitus by being called by it.

That relationship did not stay smooth. About a year before the interview, several large insurers blocked AI phone calls entirely — not because of Infinitus, but because a wave of bad ones arrived and no call center was going to train staff to tell vendors apart. Working through it produced a better outcome: dedicated call center teams for the Infinitus agent, a private phone number, shorter wait times and digital data exchange.

The human incentive behind enterprise sales

Jain's sharpest go-to-market observation is that the buyer's personal risk, not the company's ROI, is what actually gates an enterprise deal.

"I will say even to this day, every champion of ours has gotten one, if not two promotions and within the first two years of using us." — Ankit Jain, Infinitus

Infinitus tracks that internally. If someone stakes their standing on adopting an unproven technology, making them look good is part of the product.

Key lessons from Ankit Jain's playbook

1. Validate before you build, even if it is awkward. Jain's diligence calls as an investor were his customer discovery, and they produced his first customer. 2. Test the riskiest assumption with the cheapest possible fake. A Wizard of Oz call with a robotic voice answered the only question that mattered: will the human stay on the line. 3. Borrow the deployment model from an adjacent industry. A human in the loop that generates training data is how autonomy gets built, in cars and in call centers. 4. Constraints built for weak models become moats when models improve. The guardrails Infinitus needed in 2019 are what enterprise compliance teams buy in 2026. 5. Count the acceptance rate, not the rejection rate. Sixty percent adoption of something nobody had ever done was a starting position of strength. 6. Sell to the champion's career, not just the company's P&L. Track whether your buyers get promoted. 7. Product market fit is a ceiling you keep raising.

"Every year, there's a moment when you feel like you found product market fit for that stage, and then you kind of work your way into that, and then you hit some kind of a ceiling, whether it's in your mind or from a business perspective. And you have to work hard to kind of punch through that and make the ceiling your next floor." — Ankit Jain, Infinitus

FAQ: Infinitus

Q: What is Infinitus? A: Infinitus Systems is a healthcare AI company whose voice agents make and handle the phone calls behind benefit verifications, prior authorizations, co-pay assistance checks and patient outreach — work that otherwise occupies thousands of people at providers, pharmacies and drug supply chain companies.

Q: Who is the CEO of Infinitus? A: Ankit Jain, who co-founded the company in February 2019 with Shyam Rajagopalan. Jain previously helped start Gradient Ventures, Google's AI venture fund, and had founded and sold a company before that.

Q: How much funding has Infinitus raised? A: $102.9 million in total. The most recent round was a $51.5 million Series C led by Andreessen Horowitz, announced in November 2024, with Memorial Hermann Health System joining existing investors Kleiner Perkins, Coatue and GV.

Q: Who are Infinitus's investors? A: Andreessen Horowitz, Kleiner Perkins, Coatue, GV (Google Ventures) and Memorial Hermann Health System.

Q: How long are the calls Infinitus automates? A: The average automated call runs over 30 minutes and follows detailed clinical and administrative procedures. Infinitus says its agents verify answers in real time, because about a quarter of the information insurance agents give out is incorrect.

Q: When did Infinitus start doing AI phone calls? A: It made its first production calls in late 2019, about six months after the company was founded and three years before ChatGPT launched.

Sources: Listen to the Full Founder Story

Ankit Jain, Co-Founder & CEO of Infinitus — left Google's AI venture fund to automate healthcare's 30-minute back-office phone calls, validated the idea on diligence calls before the company existed, and built the guardrails that now sell it to insurers. Listen to the full interview.