Sundar Subramanian, a partner at PricewaterhouseCoopers (PwC) consulting affiliate Strategy&, has joined population health management company Zyter|TruCare as its newest CEO, Fierce Healthcare exclusively reports.
He has spent much of his career helping modernize healthcare companies, also leading PwC’s strategy, value and digital transformation business. Subramanian once co-led the Medicaid/Medicare Center of Excellence and Core Operations service for payer markets at Booz & Company and held senior roles at WellCare and McKinsey.
Now, he is in charge of running the healthcare IT vendor covering more than 44 million members, helping 45 health plans on prior authorization, as well as case, utilization and disease management.
Primarily, he joins Zyter to reshape the company’s product offerings around AI agents designed to communicate together. The company believes its AI agents, to be released within the coming weeks, are superior to services from competing vendors and will reduce the administration burden by up to 50%.
While other AI agents may improve efficiency in the area they are deployed, the fundamental division between teams remains. With Zyter’s new rollout, AI agents will complete a task and then interact together, according to the company.
Zyter was founded by billionaire owner Sanjay Govil, chairman of Infinite Computer Solutions, a digital engineering and IT services company. Govil started Zyter to focus on healthcare technology. In 2021, the company acquired Casenet, a provider of population health products, from Centene, later rebranding the company to Zyter|TruCare.
Through conversations with C-suite executives, one theme grew clear to Subramanian: technological advancement has often not led to efficiencies, particularly within healthcare. A survey from PwC found that eight in 10 companies are not realizing value from technological investments. Data systems remained siloed or based on outdated legacy technology.
“Isn’t that very startling?” asked Subramanian in an interview.
One problem, he believes, is that technology is adopted but doesn’t seamlessly fit within existing systems. Outcomes don’t improve because the new technology is layered.
Other vendors may sell licenses, but they usually “drive workflows without intelligence,” he said. The result is more revenue but backend processes stay largely the same.
Further, AI vendors may improve individual use cases, but payers are still left struggling to coordinate all products across teams and business lines, making it difficult to effectuate change. This leaves executives searching for more and more ways to improve star ratings, lower costs and foster more engaged care.
Normally, a prior authorization request is reviewed by a nurse, who manually checks eligibility and benefits, determines medical necessity and checks provider network verification. When information is missing, they must confirm missing links with providers, who are busy with patients.
If a claim is approved, the team will alert the provider and could refer the member to care management staff, though that does not always occur. Since data is often not integrated, the care management team operates off a different set of information and must verify data once again, before finally creating a plan and contacting the member. The care management nurse then enrolls the member and monitors progress.
This process is still done partially through fax, with clinical documentation up to 600 pages in length. It is completed through independent operating systems and usually includes conflicting information, said Subramanian. Patient data cannot be captured in real-time and teams cannot easily interact together, meaning nurses lack context and clinical notes. Interventions are delayed and proactive care is difficult.
Zyter’s AI agents can evaluate unstructured clinical information and summarize these same documents much more quickly and with near-total accuracy, he explained. A nurse will review the prior auth determination. If there is a denial issued by a physician, the AI agent will contact the provider and give an avenue to submit an appeal. Providers will be able to speak to the AI agent and get authorization approved, when possible, on the phone. That information is automatically synced and updated for the care management team.
If the request is initially approved, the AI agent still contacts the provider, this time with an approval letter. It instantly recognizes an opportunity for disease management intervention and begins the process of referring members to care management teams, prioritizing certain members. The agent will create the care plan based on member goals and engagement insights and conduct voice outreach to members for assessments. A nurse then reviews the plan and enrolls the member. An agent monitors the enrollee’s progress and flags concerns for nurses as they arise.
“That’s a much more proactive healthcare model that is going to be very exciting to build,” he said.
Other features of the AI bots? Machine learning to anticipate denials and auto-generate appeals, streamlined post-acute care assessments and continuous checks against regulatory requirements.
Zyter will soon offer 25 agents to improve these use cases and more. Health plans are also able to create and customize agents of their own to solve issues unique to their business.
For example, a large payer might use several systems for prior authorization processes that have been configured and updated over many years. Or, utilization management may be handled by a vendor through another system. Logistical complexities must be handled perfectly by the agents.
“You don't just configure your agents to then take actions on all scenarios,” said Subramanian. “You have to configure it to say, ‘Exclude those scenarios where another vendor is taking action,’ and then be able to accept API inputs for their decisions that the other vendor has made on behalf of the plan and accept it. There are hundreds, if not thousands, of scenarios like this that happen in a real-world environment, and the flexibility to deal with all those is what I call the last-mile problem.”
Plans can manage these new systems in-house, or insurers can use a hybrid model where Zyter manages some tasks and trains internal health plan teams on how to use the agents. Additionally, Zyter has clinical staff and can completely take over provider services and utilization management entirely.
“We think there's tremendous amount of efficiency to unlock, to reinvest in care management,” he said. “The administrative burden of the clinicians and the nurses and doctors is inordinate right now.”
AI agents are an increasingly popular solution rolled out by companies, evident at this year’s America’s Health Insurance Plans 2025 conference. AI-powered tech company Anterior CEO Abdel Mahmoud, M.D., pitched a future where prior authorization is virtually eliminated because of the utility from agents.
This year, Salesforce released Agentforce for Health and inked partnerships with athenahealth, Availity and Infinitus.ai. Commure launched its own suite of AI agents for tasks including patient navigation and care management. And Insight Health joined the ranks of startups focused on medical documentation through AI agents.
“Health plans are saying ‘We know this platform and that platform have their different bells and whistles, but our medical management operation is not doing what we need it to do,’” said John Gaines, senior vice president of marketing for Zyter|TruCare.
Frustration at insurers over prior authorization policies is as high as ever, prompting major health plans to pledge to reduce the numbers of services subject to these restrictions. Days after insurers made that pledge, the Centers for Medicare and Medicaid Services Innovation Center rolled out a new model to test new prior authorization requirements in traditional Medicare, a decision opposed by the Medical Group Management Association.