Innovaccer CEO on the Power of AI to Transform Prior Authorization

This month we are doing a series of interviews about how health plans and provider organizations are working toward reducing the drivers of friction in prior authorization processes, as the payers get to ready to comply with the CMS rule around the use of FHIR APIs by January 2027. 

I wasn't surprised to discover that one of the companies innovating in this space is Innovaccer, which describes itself as a healthcare intelligence platform. The company seems to have its finger on the pulse of several areas of health system transformation Last August, Innovaccer launched Flow Auth, an AI-powered prior authorization solution that is part of its revenue cycle performance platform.

I asked Abhinav Shashank, cofounder and CEO at Innovaccer, what he hears from provider customers about their experience with prior authorization.

“Prior authorizations are one of the most frustrating parts of providers’ workflow today,” he said. He added that he thinks the reality is that the cost structure in healthcare is out of control, and the process of utilization management and authorizations based on care guidelines is something that we need as a country. “But the frustrating part of it — where I think everyone gets stuck — is that the authorization process has become something that takes too much time. If you want people to follow a certain set of guidelines, you probably want to give them real-time inputs on whether something is allowed per a guideline or not.”

He described Kaiser Permanente as a role model here, given that they have a completely integrated value-based care model. They have created a way for physicians across the board to follow a certain set of guidelines, so there is really no need for a physician to seek a prior authorization, because all of the guidelines are present in front of the provider then and there. But that is not the case in most healthcare settings. “The thing that creates a challenge is that today if I am a physician wanting to order an imaging, that authorization could take me six or seven days to get cleared up. In the meantime, what do you want the patient to do?”

Shashank thinks the government is right to push on two things. With the need to reduce fraud, waste and abuse, the push is happening in one direction through the WISER model. In the the other direction they are pushing on health plans, saying we need to create more real-time frameworks to get physicians and providers to have access to guidelines that they should follow in real time.


“Most of our customers are value-based care champions and and they want to adhere to guidelines,” Shashank said. “But if we know those guidelines up front, and if we can use AI to process the patient context, and provide that guideline based on the patient context right then and there, that would be such a huge win.”

Shashank said the innovation to overcome this problem requires changes on the payer side as well as a modernized revenue cycle module on the provider side.

Innovaccer describes its AI-enabled prior authorization solution Flow Auth as leveraging an end-to-end automation framework powered by a seamlessly integrated ecosystem of AI agents. It streamlines relevant stages of the PA process, from eligibility verification and submission to appeals, while adhering to payer-specific guidelines. Flow Auth:

• Detects when authorization is required earlier in the scheduling and financial clearance path
• Assembles payer-ready packets by pulling the right clinical context and documentation signals
• Validates completeness and medical necessity indicators before submission
• Tracks status and escalates exceptions so cases do not fall into silent delays
• Feeds denial prevention by reducing avoidable “no auth/insufficient documentation” outcomes, improving first-pass yield, and tightening days in accounts receivable
• It also can generate appeal letters, with intelligent exception handling that keeps humans in the loop.

Shashank described the cost structure in creating prior authorization requests on the provider side and how Innovaccer has worked to bring that down. “In creating a prior authorization request, in terms of human time, estimated in dollar terms, it was somewhere in the range of $40 to $60, which is effectively 15 to 20 minutes of time for whoever is creating that prior authorization, plus some physician time. That $40 to $60 adds nothing to care,” he said. “We have one of the most comprehensive data infrastructures that exist in the country, and we built out the AI framework to create prior authorizations, and it led to an 83% reduction in time, and the cost went from $43 to $7 and this is not even using full AI capabilities that are fully autonomous. I think that time is going to get reduced dramatically.”

Shashank added that you might think if something is costing less, the quality would have reduced. “But here, with things like the first-pass completion, it actually jumped from 60% to 83% so more prior auths were getting approved in the first pass, and the time it took for payers to approve it also reduced dramatically because they weren't going back and forth on all the data that they needed to approve a prior authorization,” he explained. “We launched this in beta only to our top customers initially, and we have had glowing feedback.”

He sees AI playing a massive role in solving this problem. “I think if we remove this burden, healthcare providers and probably payers as well, are going to see a reduction of tens of billions of dollars.”

Platform AI vs. point solution AI

Innovaccer has just released a report on the state of AI deployment more broadly. The company said that input from healthcare decision makers indicates that healthcare has reached a turning point where AI is measurably reducing administrative waste and access delays, but only a small fraction of systems have the data infrastructure needed to scale those gains equitably, raising concerns about a widening digital divide.

One of the things the report says is that the most important trend for 2026 is platform AI vs. point solution AI. I asked Shashank if he could elaborate on that point. 

He said when any new technology arrives, point solutions proliferate. This happened with the widespread adoption of EHRs, which has forced the industry to try to solve for EHR interoperability. “What CIOs should think through is that if you are worried about EHR interoperability today, then think about what agentic interoperability will be like,” Shashank said. There might be 10 leading EHRs and maybe 50 more EHRs, and there might be 200 point solutions. But there might be 2,000 AI point solutions that are proliferating now. “How do you get these things to work together at a later point of time? That is an IT nightmare that no one is ready for. In agentic workflows, all of the exhaust of these agents, is unstructured data. So how do these things work together? If your prior authorization solution is working differently than your coding solution, and your coding solution does not talk to your scribe solution, and your scribe does not talk to your contact center, are we getting ourselves  into a mess?”

EHR interoperability cost the system billions of dollars at a national level, but Shashank thinks that agentic interoperability is going to be far more expensive if you don't solve it structurally today. “That is why our top customers, including KP, Ascension and others, are thinking about these things from a platform perspective — how do I set up a modern data infrastructure that allows for these things to talk to each other?”

Shashank stresses that AI is now at a point where it's inevitable. “We are not going to have a health system that doesn't have AI. The difference between winners and losers is going to be whether you had a common infrastructure that got all of these things together, and a data infrastructure that allowed you to do all of these things in a common framework and move quickly ahead of everyone else,” he said. “I don't think the EHR is the answer either because you're talking about a common data infrastructure and intelligence infrastructure, you're not talking about a record structure. The records that are being created by the EHR are such a small component of all of the data that you are going to create.”

If a CIO does not have a strong data infrastructure built out today, they are not going to succeed over a longer term, he said. “For most of our customers, it's great that you want an AI strategy, but the first step of the AI strategy and autonomy is having a strong data infrastructure that is interoperable with all of the systems. Because if you don't have an interoperable data system sitting in the middle, you will get into trouble sooner or later.”