Building a High Growth Vertical AI Company in Healthcare: Ganesh Padmanabhan, CEO of Autonomize AI (Part 6)

Sramana Mitra: It sounds like in your go-to-market strategy, you’ve operated as a co-pilot to human in the loop. Is that a correct observation?

Ganesh Padmanabhan: That is correct, yes.

Sramana Mitra: That’s astute, because I’m hearing from all the investors right now that they want to make sure it’s human in the loop, because people don’t want full automation; they want human augmentation.

Ganesh Padmanabhan: Yes. It is not that they don’t want it, there’re two different problems. One is, it’s impossible to do that in a healthcare context or any regulated industry. In an agentic workflow, basically you’re starting with the premise that all the data and the connectors and the things you need to go perform their entire workflow exists in a digital form, which is often not the case in healthcare.

I’ve met utilization management nurses who’s got a 4,000- page binder on their desk, and they review that every time a case of a certain kind comes in. That knowledge is not digitized yet.

Sramana Mitra: Digitization has not happened yet.

Ganesh Padmanabhan: It hasn’t happened. One of the things we do is capture that nuanced tacit knowledge as a copilot. We prompt them to say why did you make that decision, even though the AI said it’s recommended for this thing. They add a couple of comments in there, from which we learn. Then over time, we see these copilots turning into autopilots. We’re not there yet.

The other problem, is it’s a regulated industry. CMS has mandated that you cannot deny a claim or a medical procedure through AI. It’s much harder to go in and be a copilot because you’vee to understand the workflows, understand and engage the users in there. That’s how we build the moat, right?

Early on, we said that we’re not going to agent wash for them. We’re going to be the best copilot. Now, underneath the covers, we’re using autonomous agents to do parts of that.

Sramana Mitra: So what agent technology did you use?

Ganesh Padmanabhan: We use a broad set of capabilities. We have a library of large language models, small language models, curated pre-trained models, like Bard-based models for doing specific tasks. We use auto gen and several other agentic frameworks, if you will, like LangGraph. We use that as a starting point, but we’ve built our own multi-agent orchestration framework.

Sramana Mitra: Okay.

Ganesh Padmanabhan: So, we have a platform called the Autonomize genesis platform. All the co-pilots are applications that are composed of multiple agents within the platform.

We have launched and deployed about 50 agents with customers. Each of those agents can be combined in different combinations to deliver a particular copilot.

Sramana Mitra: But it’s your technology,

Ganesh Padmanabhan: That’s our core technology. We use standard off-the-shelf LLMs, wherever it’s appropriate. We use the GPT 4o for Azure based deployments. We use the Google Gemini, and Gemma for GCP based deployments and so forth. We use Anthropic for AWS deployments, but wherever we need to use the public LLM, we actually use that. But then most of the other orchestration and the compound AI orchestration is done using our own proprietary technology.

Sramana Mitra: I think the state of the union currently is, it’s a lot of system integration that you have to do if you want to build a solution. It’s a humongous system integration problem that you’re solving.

Ganesh Padmanabhan: It’s interesting you say that and a very astute observation. I would characterize it this way. AI has a lot of ingredients, not enough recipes. What we are really doing is take these building blocks and get to value. In the process of doing that, the problem with system integration is, it’s a throwaway work. Every single time you’re integrating it, you’re doing this over and over again.

Sramana Mitra: Yes. It’s productized system integration.

Ganesh Padmanabhan: Productizing system integration is a big part of what we do.

Sramana Mitra: The only way you can build scalable companies like this – it’s a vertical AI company that you’re doing – is by doing that kind of productized system integration that is specific to a problem and the workflow and everything. So it’s perfect.

Ganesh Padmanabhan: Exactly. The other longer term capability we are building is the knowledge graph, because today, no data exists. That takes insights and context from one workflow within a health enterprise and uses in the other, because the tools don’t talk to each other.

The current state of the art is not there. We believe we’ll build the most unique data set in the world that has the ability to build a healthcare AGI that does everything that is needed to go solve an entire thing like the future health enterprise should be. Kaiser Permanente should be able to launch a new Medicare Advantage plan or a value-based care plan with two doctors, 10 nurses and one platform, or one API from Autonomize AI where all the underlying operations of running the administrative processing are done by agents underneath the covers. But the only way you can get there is by having the combined context across workflows.

Sramana Mitra: Yes. You’re in the process of building that. So go back to what you said about getting in with the first use case and then the second use case, and then getting these multimillion dollar contracts from the same customer.

What is the lifecycle of your first use case from a dollar value point of view or ARR point of view, and then how does the dollar value progress in these accounts?

Ganesh Padmanabhan: It’s a good question. So, we use a very transparent, simple pricing model. That’s a huge part of what we do, because enterprises like predictability and they like to align to value. So that’s two big things that we want to do.

For that, we use a couple of techniques. One, our usual pricing model includes a technology platform fee. And then we have per transaction and per use case pricing for agents and copilots that are used for them.

It’s very transparent. They get to value. We don’t demand them to start at a couple of million price point. In fact, we have folks starting at $150K as the initial ARR. We try to do a combination. There’s a consultative selling aspect to get the customer comfortable and get started on this particular journey. Right after you do that, you have to flip over to a product-led motion to drive the adoption because our entire business model is land and expand.

Now, we are evolving beyond that. We are telling large enterprises to do a broader engagement where we can look at how do they transform those five departments in 24 months. So, we’re entering in those that phase in our phase of growth.

But initially it was all about giving them bite-size value. You have to solve that problem so well that you get them hooked. From that moment on, it’s the product pulling the adoption rather than sales pushing the adoption. I think we spend a lot of time getting that right. I think we’re still work in progress, but I think we’ve really solved a problem.

After you solve one or two use cases, we enable a playground with all the 50 agents on our platform free of cost for the customer to go start experimenting their new use cases. What that led to is them coming to us with a roadmap of 20 things they want to build with us over the next two years. Ten of them may not make any sense, but then we are building a deeper relationship. They’re seeing value and want to do more with us.

Sramana Mitra: How many health systems do you have?

Ganesh Padmanabhan: We have about 15 health systems in the health enterprises that include health plans and life sciences companies.

Sramana Mitra: It’s not just health plans right now?

Ganesh Padmanabhan: All these lines are blurred. We just announced a strategic partnership with Altais, a large specialty and primary care network health provider. They’re turning into a payvider, or large providers who are complex enough to take on at-risk patients for an insurance company. They call it delegate plans. They have to be the caregiver and also the care administrator. Companies like them have structures and systems to do the healthcare delivery like EMR systems, but everything else they need to do on the health plan side of the house is very manual.

We just announced a large partnership with them. It’s a commercial relationship where they will deploy our AI copilots across multiple workflows across their entire spectrum. They’re also partnering with Abridge for their EMR side.

So, we’ve a combination of mostly health plans, Medicare advantage plans, national health plans, and regional health plans. We’re now getting into what we call payviders. These are large providers who are complex enough to take on at-risk patients like the value-based care organizations.

This segment is part 6 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