Skip to main content

Agentic AI Raises the Stakes for Healthcare Software Leaders

May 13, 2025
Image: [image credit]
ID 132056548 © Wrightstudio | Dreamstime.com

Victoria Morain, Contributing Editor

Rahul Sharma, CEO of HSBlox, presents Agentic AI as the natural evolution of healthcare technology. In his Q&A with HIT Leaders & News, Sharma argues that autonomous agents, systems capable of interpreting, deciding, and acting in real time, will make traditional SaaS platforms obsolete. He frames Agentic AI as the infrastructure behind future-ready h,ealthcare, enabling smoother care transitions, proactive claims management, and more efficient clinical workflows.

It is a compelling case. But that vision, while technologically feasible, leaves major structural and regulatory questions unresolved. For vendor leaders and provider CIOs, Sharma’s optimism should be matched with a sober assessment of what it will actually take to implement, govern, and fund autonomous decision-making in operational healthcare environments.

Agentic AI Raises Expectations That Outpace Most Healthcare IT Environments

Sharma positions Agentic AI as a shift from static data platforms to dynamic operational engines. Rather than just generating insights, these agents are designed to act. In theory, that means reducing administrative burdens, speeding up decisions at the point of care, and eliminating inefficiencies across systems.

Putting that into practice requires real-time interoperability, standardized data flows, and organizational trust in autonomous action. Most health systems are not built for that. Agentic deployment assumes readiness in areas where current workflows remain fragmented or manually driven. Governance, not just code, becomes the central operational barrier.

This shift will also upend traditional performance metrics. If agents make autonomous decisions, health systems will need new oversight models to track whether outcomes improve or degrade. Vendors offering Agentic capabilities must be ready to deliver not only functionality but also clear metrics tied to compliance, efficiency, and risk reduction.

Reimbursement and Revenue Pathways Remain the Critical Bottleneck

One of the strongest use cases Sharma outlines is chronic disease management. Agentic AI can support automated outreach, data collection, and escalation protocols across multi-team environments. But none of that translates into scale unless the underlying payment system supports it.

Current Medicare billing codes for chronic care management rely on time-based documentation by care team members. AI agents do not currently qualify under CPT logic. CMS has not yet defined how automation interacts with reimbursement, and commercial payers have shown no urgency to build coverage for agent-based workflows. This gap will stall deployment for any organization dependent on fee-for-service revenue or hybrid models.

Enterprise health systems considering AaaS models should press vendors to explain how outcomes from agent-driven workflows can be documented, attributed, and billed. Otherwise, ROI will be confined to efficiency gains alone, which can be difficult to quantify or justify to boards during procurement cycles.

Clinical Safety and Regulatory Compliance Must Be Proactively Addressed

Agent-driven care coordination, as described by Sharma, highlights real opportunities to reduce communication gaps and handoff failures. But agents that take action on clinical data or intervene across care settings introduce new forms of clinical risk.

Today’s AI governance structures are designed around decision support. Agentic systems move beyond support and into autonomous orchestration. This raises liability questions, documentation challenges, and compliance gaps that neither FDA nor ONC has resolved. Without a clear framework for auditing autonomous actions or certifying agent-based interventions, health systems expose themselves to new types of regulatory scrutiny.

Sharma presents Agentic AI as a complement to the clinical workforce. But as autonomy increases, human oversight must be tightly integrated into the design. Vendors will need to build transparency into every decision an agent makes, and provider organizations will need escalation protocols for when agentic decisions misfire or contradict human judgment.

Integration and Infrastructure Constraints Still Determine Viability

Agentic AI systems rely on rapid data access and seamless communication between disparate platforms. Sharma emphasizes that unlike traditional SaaS, these agents can retrieve and act on data without pre-built pipelines. The challenge is that most health systems are still fighting to get consistent EHR access, normalize data across vendors, and manage privacy restrictions that throttle real-time exchange.

This means that vendor roadmaps must go beyond promising autonomy. They must offer practical implementation models that acknowledge the messy, inconsistent digital ecosystems most clients still operate in. For many health systems, early adoption will depend less on the technology itself and more on how well it can be deployed in parallel with legacy infrastructure without introducing clinical or financial risk.

For healthcare executives, the strategic question is not whether Agentic AI is the future. The question is how to prepare for a future where software does not just inform human decisions but initiates them. That shift requires investment in oversight models, data governance maturity, and reimbursement advocacy. Vendor partners will need to shoulder more responsibility for outcomes and alignment with institutional risk tolerance.