Sundar Subramanian, CEO of Zyter, is an experienced strategy and healthcare executive with a background spanning management consulting, digital transformation, and large-scale healthcare operations. Before becoming CEO in July 2025, he spent more than 16 years at PwC’s Strategy&, where he led the U.S. strategy consulting business as well as its Enterprise Strategy, Value and Digital Transformation practice, advising major organizations on growth, operating-model redesign, and AI-enabled transformation. Earlier in his career, Subramanian co-led Booz & Company’s Medicaid and Medicare Center of Excellence and worked extensively with healthcare payers, following senior roles at WellCare and McKinsey & Company. At Zyter, he is leading the company’s expansion around agentic AI and intelligent workflow orchestration, with an emphasis on using AI to redesign end-to-end business processes rather than simply automate individual tasks.
Zyter is a privately held enterprise technology company with more than two decades of experience in digital healthcare, where its TruCare population health platform has established a footprint supporting more than 44 million lives across over 45 healthcare organizations. The company is now building beyond its healthcare roots with an AI execution platform designed for complex and highly regulated workflows across healthcare, government, and financial services. Its architecture combines Zyter Symphony, which coordinates AI agents, enterprise systems, and human actions; Zyter Praxis, which packages agents, rules, data, and human oversight into end-to-end workflow modules; and the TruCare Digital Core, which connects these workflows to systems such as electronic health records, claims platforms, enterprise resource planning systems, and customer relationship management software. The broader goal is to make agentic AI operational within existing enterprise environments while maintaining governance, auditability, and human oversight over consequential decisions.
You spent more than 16 years at Strategy&, leading its U.S. strategy consulting business and working extensively in healthcare, after earlier experience at WellCare. What convinced you that healthcare’s next major leap would come from agentic AI and workflow execution rather than another generation of traditional software?
Throughout my career, I’ve had the opportunity to see healthcare from several perspectives, as an operator, a strategy consultant, and now as someone building technology. Across those experiences, one observation became increasingly clear. Healthcare wasn’t struggling because it lacked technology. It was struggling because technology wasn’t changing how work got done.
Over the past decade, the industry invested heavily in digitization. We modernized infrastructure, expanded analytics, improved interoperability, and more recently embraced AI. Those investments created important new capabilities, but they didn’t solve one of healthcare’s most persistent challenges: coordinating work across people, systems, policies, and organizations. Too often, technology was layered onto existing operating models instead of changing how they functioned.
That realization changed how I think about transformation. The next phase isn’t defined by deploying more AI. It’s about redesigning how work gets done. AI is an extraordinary enabler, but its greatest value comes from helping organizations execute more effectively by orchestrating work across people, systems, policies, and AI.
That’s what drew me to Zyter. After years of helping organizations develop transformation strategies, I wanted to help build the execution layer that turns those strategies into measurable outcomes. For me, the future of healthcare isn’t about making AI more intelligent. It’s about helping healthcare operate more intelligently.
Zyter positions itself as an execution layer for the agentic enterprise, with Symphony orchestrating agents and humans, Praxis executing workflows, and TruCare grounding them in enterprise data. What can this architecture accomplish that individual AI copilots cannot?
Enterprise transformation has never been limited by a lack of ideas or intelligence. Organizations usually know what they need to do. The harder challenge is consistently translating those decisions into action across people, systems, policies, and workflows.
That’s where I see the next phase of enterprise AI. Individual AI assistants make people more productive. Enterprise AI should make organizations more effective. It should coordinate work, execute workflows, adapt as conditions change, and continuously improve how the organization operates.
Our architecture reflects that philosophy. Symphony orchestrates AI agents, enterprise systems, and human expertise. Praxis executes workflows, and TruCare provides the operational foundation that grounds those workflows in trusted enterprise data. Together, they create an execution layer where AI becomes part of how work gets done rather than another tool people use alongside their work.
Organizations will ultimately measure AI by its ability to improve outcomes. Productivity matters, but long-term value comes from helping enterprises execute more consistently, adapt more quickly, and learn from every workflow they run.
Agentic AI becomes much more complex when multiple agents are sharing data, handing off tasks, and escalating decisions to humans. What has Zyter learned about making multi-agent systems reliable enough for healthcare at scale?
As AI moves beyond individual use cases and becomes part of everyday operations, the challenge changes. It becomes less about how intelligent any one agent is and more about how work is coordinated across systems, policies, people, and AI.
We’ve found that specialization matters. Healthcare doesn’t rely on one person to do everything, and AI shouldn’t either. Agents that have clearly defined responsibilities are more reliable, easier to govern, and easier to improve over time. The orchestration layer coordinates those agents, manages handoffs, applies policies consistently, and determines when human expertise should be brought into the workflow.
Visibility is equally important. Teams need to understand how decisions were made, what information informed them, and how they contributed to the outcome. Capturing decision lineage creates that transparency while allowing every workflow to become a source of learning and continuous improvement.
Reliable AI isn’t defined by how many agents are deployed. It’s reflected in how consistently work gets done, how confidently people can trust the results, and how effectively the system improves over time.
Healthcare organizations want AI systems to improve over time, but they also need predictable behavior and auditability. How do you enable continuous learning without allowing systems to drift away from clinical guidelines, payer policies, or approved workflows?
I don’t believe every decision should be automated through AI. In healthcare, there are many decisions where deterministic logic, established business rules, or clinical policies can provide a clear and consistent answer. Where that’s possible, we should use them rather than introduce unnecessary uncertainty.
The more interesting opportunity is combining those deterministic approaches with AI. We’ve successfully deployed neuro-symbolic logic that creates a verification layer around AI-driven reasoning. AI can interpret complex or unstructured information, while deterministic rules and symbolic logic verify that the resulting action remains consistent with established policies, guidelines, and approved workflows. That reduces uncertainty and improves accuracy compared with relying on an AI-driven approach alone.
This also changes how I think about continuous learning. The goal isn’t to give AI greater freedom to make decisions over time. It’s to learn from execution while preserving the controls that determine how consequential decisions are made. You can improve how the system interprets information, handles exceptions, and coordinates work without allowing it to drift away from the rules that govern the underlying process.
That combination of AI reasoning, deterministic decisioning, and continuous verification is how you create systems that can improve over time while remaining predictable, auditable, and accountable.
Zyter has been exploring C-RLM, a recursive approach for synthesizing long and fragmented clinical histories. How important will long-context reasoning become in healthcare, and could agents eventually maintain a continuously evolving understanding of each patient?
Healthcare has never struggled with a lack of information. The challenge is that a patient’s story is spread across years of physician notes, lab results, medications, imaging, claims, and care interactions. Making sense of that history requires more than processing a large amount of data. It requires connecting evidence, recognizing patterns, and building a coherent understanding of each patient’s journey.
Long-context reasoning will become increasingly important as AI moves beyond assisting with individual tasks and begins supporting longitudinal care. Our work with C-RLM reflects that direction. The objective isn’t simply to summarize more information. It’s to synthesize fragmented clinical evidence into a structured understanding that remains grounded in the underlying record. Every conclusion also needs to remain traceable to its source so clinicians can understand how it was reached and validate it against the evidence.
As these capabilities mature, AI agents will increasingly help maintain a continuously evolving understanding of each patient’s journey. That understanding should become richer as new information becomes available while remaining transparent, evidence-based, and grounded in trusted clinical data. The role of AI is to give clinicians a more complete picture so they can make better-informed decisions with greater confidence.
Zyter is participating in the CMS WISeR program, where AI can support workflows while licensed clinicians retain decision-making authority. Is that roughly where the autonomy boundary should sit today, and what would need to change before AI agents could take on more consequential decisions?
I don’t think there’s a single autonomy boundary for healthcare. The appropriate role for AI depends on the workflow, the level of risk, and the confidence organizations have that the system will behave consistently under real-world conditions.
Programs like WISeR reflect that reality. AI is already very effective at synthesizing information, applying established policies consistently, coordinating administrative workflows, and reducing repetitive work. Licensed clinicians continue to provide judgment for decisions where context, ambiguity, and accountability matter most. That’s an operating model that reflects the strengths of both AI and human expertise.
Over time, those boundaries will continue to evolve, but they should evolve because trust has been earned, not because models have become more capable. Confidence comes from evidence, transparency, validation, and consistent performance in production. Every expansion of AI’s role should be supported by the same standards healthcare applies to any other critical capability.
I don’t think success will be measured by how autonomous AI becomes. It will be measured by whether patients receive better care, clinicians have more time to focus on complex decisions, and healthcare operates more effectively as a result.
Zyter recently moved TruCare onto a cloud-native AWS foundation to support AI orchestration at enterprise scale. How much of the challenge in agentic AI is about the models themselves versus the infrastructure connecting data, systems, policies, and people?
Foundation models are advancing at an extraordinary pace, and that’s good for the entire industry. The conversation is gradually shifting from what models are capable of to how they operate inside real enterprises.
Connecting data, systems, policies, and people is where transformation either succeeds or stalls. AI has to work within existing operational environments, execute across multiple systems, apply policies consistently, and perform reliably at production scale. Without that foundation, even the most capable model remains disconnected from the work it’s intended to improve.
Our move to a cloud-native AWS architecture reflects that evolution. It provides the resilient, scalable foundation needed to orchestrate AI across enterprise workflows while maintaining the reliability, governance, and performance healthcare requires. The infrastructure is important because it enables AI to become part of how work is executed every day, not simply another application running alongside existing processes.
The long-term opportunity isn’t defined by better models alone. It comes from combining advances in AI with the infrastructure and operating model required to execute work consistently at enterprise scale.
Zyter is also positioning its execution model for other regulated industries, including financial services and government. Which parts of the platform are truly horizontal, and which advantages come specifically from Zyter’s experience in healthcare?
Regulated industries are often viewed through the lens of what makes them different. I’ve found it’s equally important to understand what they have in common. Whether you’re delivering healthcare, managing financial risk, or supporting government programs, success depends on coordinating complex work across people, systems, policies, and data while maintaining accountability for every decision.
That’s where I see the execution model as fundamentally horizontal. Orchestrating AI, enterprise systems, and human expertise, applying policies consistently, coordinating work across multiple systems, and maintaining transparency are challenges that extend well beyond healthcare. The workflows differ by industry, but the need for reliable execution is remarkably consistent.
Healthcare didn’t change what we built. It raised the standard for how we had to build it. Clinical workflows demand precision, interoperability, governance, resilience, and trust because the consequences of getting it wrong are significant. Designing for that level of complexity created an execution model that can be adapted to other regulated industries where operational discipline, compliance, and accountability are equally critical.
As AI becomes more deeply embedded in enterprise operations, I think the conversation will shift away from whether a capability was built for a specific industry and toward whether it can execute reliably in complex, regulated environments. That’s the standard we’ve designed for from the beginning.
Zyter’s Rural Health Orchestrator brings together virtual care, remote monitoring, care management, and clinical services. What can agentic orchestration solve in rural healthcare that telehealth alone cannot?
Telehealth solved an important access challenge. The next challenge is ensuring that every interaction leads to coordinated action and measurable outcomes.
Rural healthcare has made significant progress expanding access through telehealth, remote monitoring, and broadband investment. Yet outcomes have not kept pace because too many digital interactions remain disconnected from the broader care journey. A virtual visit that doesn’t update a care plan, trigger follow-up, or coordinate the next step improves access, but not necessarily health.
That’s where orchestration becomes essential. It connects telehealth, remote monitoring, care management, community resources, and clinical workflows into a single, coordinated system so information flows seamlessly and action follows automatically. Every interaction becomes part of a continuous care journey rather than an isolated event.
The opportunity isn’t simply to expand access. It’s to convert access into outcomes. When care is connected across people, systems, and workflows, clinicians spend less time navigating fragmented processes, patients receive more consistent follow-up, and rural health systems can deliver better outcomes with the resources they already have. That’s where I believe AI orchestration creates its greatest value.
With the new PwC collaboration, what would distinguish a successful AI transformation from another healthcare AI pilot? Which measurable outcomes would convince you that AI has fundamentally improved how a payer or provider operates?
Healthcare has spent years investing in digital capabilities. Today, most organizations have access to data, AI, and modern technology. The next phase isn’t about adding more capabilities. It’s about redesigning how work gets done.
That’s why I don’t measure transformation by the number of AI models deployed or pilots launched. I look at whether an organization’s operating model has changed. Are decisions moving faster? Are workflows more coordinated across departments and systems? Are administrative burdens decreasing while clinicians and care teams spend more time on activities that improve patient care? Are those improvements repeatable across the enterprise rather than isolated to a single use case?
That’s the philosophy behind our collaboration with PwC. PwC brings deep expertise in healthcare strategy and transformation. Zyter provides the execution layer that operationalizes those strategies across complex workflows, connecting AI, enterprise systems, policies, and people so organizations can consistently turn strategy into action.
For me, successful AI transformation happens when AI is no longer viewed as a separate initiative. It becomes part of the operating model. At that point, organizations aren’t measuring the success of AI itself. They’re measuring better outcomes, lower administrative costs, more coordinated care, and a healthcare system that operates more effectively because execution has fundamentally improved.
Thank you for the great interview, readers who wish to learn more should visit Zyter.
