Building a High Growth Vertical AI Company in Healthcare: Ganesh Padmanabhan, CEO of Autonomize AI (Part 4)

Sramana Mitra: Healthcare is big, but it’s cumbersome. It’s one of the biggest exit barrier systems, right? You’ve all these legacy systems that refuse to exit, that refuse to budge, and entrepreneurs have a really hard time getting their toehold – not even foothold – toehold into those homes.

Where did you find the entry point of something that you could sell?

Ganesh Padmanabhan: Everything looks very complex until you break it down to the basics, and it starts fitting in. If you really think about it, the last 10 years, health tech really exploded for investors and startups. But if you ask the average patient, has healthcare gotten better or worse, most people would say it’s gotten worse.

The reason for that is the way we solve problems in healthcare and the way we have historically solved problems. Healthcare is one of the oldest industries in the world. The labor pool allocation is done in different groups and different silos. You have a health enterprise, like you have doctors, medical assistants, nurses, prescribers, etc. All the workflows in digital health happened around empowering each of them to do exactly what they were doing, just a little better. It was just incremental.

If you have to really change the shape of healthcare, if you were to change the economics of healthcare and make a real impact, you’ve to get out of the siloed thinking and solve something monumental.

The way we looked at it was like this. Across payers, providers, life sciences, and pharma, administrative burden is one of the peak problems across the spectrum, right?

This leads to the cost of care and the access to care problems. The single biggest contributing factor for that was the fact that the data in healthcare is highly unstructured and highly contextual. For the same electronic medical record, when reviewed by a utilization management, a prior auth nurse and a health plan, they’d look for a very different context than a clinical researcher recruiting that patient for a clinical trial.  

I’ll break it down. So we realize that if you can solve this in a way where you can separate that content layer and the context layer and dynamically pair that, then you can have one platform that can go across healthcare and solve multiple problems. So that was the premise.

Now, to your question on foothold and toeholds, we spent the first year in a lot of experimentation and not so much building. We were trying different things to get in and then we found our initial toehold, which is now a foothold. It’s almost a tree trunk that’s grown and widely scaled across health insurance companies or health plans.

They have the biggest economic incentive to solve for healthcare inefficiency. Almost every healthcare professional like doctors and nurses are almost immediately only used for administrative purposes.

It signals a big pain point rather than just a nice to have vitamin. So, we went after that segment of the market, and then we came upon a problem that is controversial, highly regulated, and hard to solve – prior authorization.

For the audience, I will explain prior authorization. Before prescribing anything that is going to cost a lot of money or even a little bit of money, every doctor in the United States needs to get a pre-approval from your insurance company, ensuring that’s the right course of care. A lot of people say that it’s an unnecessary step. It’s because the incentive structure for the doctors and the providers is very much fee-based. You need to have some kind of system to make sure that the right care is being provided for the right patient at the right time. Now, the problem is that for that to happen, the doctors will compile about six months of medical history, all kinds of stuff from your electronic medical records, fill up a form, sometimes handwritten, and fax it over to a health insurance company.

The health insurance companies employs someone who is paid $20-25 an hour to type up this information into a system and attach all the attachments. Then, it goes to a nurse who’s often paid $60-70 an hour to review this and adjudicate that against guidelines manually. Finally, it gets passed on to an MD. In the state of California, you cannot deny or approve a claim or a prior authorization without an MD signing off on it. The MD is paid $250 an hour to write that up the decision.

The whole thing takes about 40-50 minutes per request. It uses all these expensive resources, the patient’s waiting for one or two weeks to hear back, and there’s often no guidance on why it’s rejected or approved, or what information needs to be furnished. This whole thing was a mess.

But nobody wants to touch it because it’s very controversial. You don’t want to be empowering just the payer or the provider. We focused on one very simple problem there – reduce the amount of time that your most expensive person will have to spend in getting an approval or a denial or for making a decision. We targeted AI in a way to be a co-pilot for them to solve that problem.

Every health plan in the in the world has this problem. We found that it’s a big problem. It’s regulated. If you’re doing Medicare or Medicaid, which is 60% of the United States health insurance, it’s a mandated requirement. It’s a validated painkiller problem. There were a lot of legacy established players who would just do it the same way it is. So, we found a toehold or rather a foothold. We got a lucky break where we convinced one customer to take a chance on us to do a very simple low-cost pilot.

We far exceeded their expectations of what we could do. We go very clearly with the story of saying that we’re solving prior authorization because that’s your biggest problem, but we don’t want to be a prior authorization company. We’re not a prior authorization company. What we are is the company that allows you to go build a scalable AI foundation to change the operating system of delivering and care for healthcare.

So how do you run a health enterprise in the most efficient fashion by curating the right data context, content and insights to power any team member across the entire spectrum?

We were very clear on that mission.

Sramana Mitra: Was this first customer a health insurance company?

Ganesh Padmanabhan: Correct. Before we started with health insurance, we had life sciences customers. We were experimenting. One thing I learned is that, like you said, if you build it, they will not come. You have to validate and make sure you’re building the right thing. Our first customer was a life sciences company. We told them, “We haven’t built anything, but here’s a prototype of what this can be. If this solves a problem for you, it’ll be huge.”

We’d validated that it will solve a real problem of reviewing medical records to match patients to clinical trials. It’s extremely important for a clinical research organization.

We said, “If we solve this for you, how much is this worth?” And they actually put a number to it. We inked a contract, and then we had the team build out the solution for them. They didn’t have to pay anything until we delivered. But we had a contract that said they will pay us if we solve a problem; and they paid us for it.

That built a foundation of what we did. Then we took the same chart review capability for patient charts into other use cases like health. The real product-market fit happened when we went into health plans and health insurance companies. But the first customer was a different one.

This segment is part 4 in the series : Building a High Growth Vertical AI Company in Healthcare: Ganesh Padmanabhan, CEO of Autonomize AI
1 2 3 4 5 6 7