The rapid rise of agentic artificial intelligence, systems capable of independent and complex decision-making, has the potential to usher in a new era for health systems, payers and life sciences companies. From automating claims processing to optimizing clinical workflows to speeding up target molecule discovery, agentic AI has the ability to deliver impressive gains in efficiency and scale.
Many healthcare organizations began agentic AI pilots in limited production environments within the past year, with some expanding from single-department proofs of concept to system-wide trials as performance and acceptance have improved.
In these early pilots, prior authorization agents review clinical notes and lab results, compile the required paperwork, communicate directly with payer systems, negotiate clarifications, and, when necessary, escalate unusual cases to human reviewers. Routine requests can be handled from start to finish within minutes, reducing delays for patients and eliminating repetitive administrative work for clinicians and health plans. Yet, as these technologies move from controlled pilots to real-world production, they will also reveal a host of challenges and unintended consequences that demand attention.
Navigating unintended consequences
Agentic AI’s ability to act autonomously introduces risks that are both technical and human. Biases in training data can result in inequitable care or coverage decisions, while opaque algorithms may make it difficult for clinicians and administrators to understand or challenge AI-driven recommendations. Operationally, even minor errors in automated processes can ripple through complex systems. For example, if an agentic prior-authorization bot misclassifies a chemotherapy regimen as “elective,” it can automatically deny coverage, trigger a downstream halt in scheduling, and leave oncologists scrambling to override the decision before a patient’s treatment window closes.
More than 250 health AI-related bills have been introduced this year across 46 states. To date, 17 states have enacted 27 of them, but no single, comprehensive federal law governs healthcare AI. With regulatory frameworks at both state and federal levels still in flux, organizations face new compliance and reputational risks.
An opportunity for command centers
In light of these challenges, could command centers offer a solution? Instead of relying solely on oversight committees, periodic audits or placing humans in the loop for every autonomous agent, a command center could serve as a centralized hub where multidisciplinary teams monitor, interpret and intervene in agent-driven workflows in real time.
Healthcare organizations have leveraged command centers for decades, from optimizing surgical throughput to coordinating patient transfers across facilities, but they could now broaden their capabilities by incorporating new AI-centric skill sets. This would enable end-to-end workflow monitoring powered by multiple agents across multiple systems. As AI agents expand across business units, systems and data sources, healthcare organizations will need multidisciplinary teams to monitor them in real time. Command centers could be an elegant design for this future reality.
The practical implementation of these centers will require tailored approaches for payers, providers and life sciences organizations. For payers, this might mean a team that oversees automated claims adjudication, quickly spotting and investigating anomalies that could signal bias or systemic errors.
In provider organizations, a command center could supervise AI-powered clinical decision support or patient flow, bringing together clinical, technical and ethical expertise to review recommendations in real time. For life sciences companies, a dedicated oversight team could monitor agentic AI platforms driving autonomous drug discovery or trial operations, intervening when the system’s results deviate from expected scientific protocols.
The possibilities for command centers extend beyond traditional operational monitoring. Some organizations could consider specialized centers focused on patient safety, revenue cycle management or population health. Each would provide an extra layer of human judgment and accountability where AI is making critical decisions.
A call for thoughtful implementation
As agentic AI moves from concept to reality in healthcare over the next decade, leaders have an opportunity to proactively shape how these technologies are developed, monitored and governed. Command centers represent one promising approach, but their effectiveness will depend on thoughtful implementation and ongoing evaluation.
To that end, healthcare leaders should first map the points in their workflows where autonomous software is expected to make high-stakes clinical or financial decisions. If the benefits outweigh the risks, they should then identify where an extra layer of human review would add critical oversight.
Launching limited-scope pilots, such as a command center focused on a single revenue cycle or care coordination task, can reveal implementation challenges before a broader rollout. Throughout these trials, clinicians and technologists need to evaluate outputs together, ensuring that safeguards are both technically sound and patient centered. Oversight models must then remain flexible, adjusting as agentic capabilities mature and regulators clarify expectations.
The full potential of agentic AI in healthcare will depend on how well we anticipate and address its risks. Command centers won’t eliminate every risk, but they can provide healthcare leaders a practical way to keep human expertise involved in AI-driven care.
Some may worry that command centers could add new layers of bureaucracy and cost, but well-designed models will streamline oversight rather than slow it down.
As pilot activity accelerates, organizations that create command centers in the near term will be best positioned to scale future innovations without sacrificing accountability or patient and clinician trust.
About the author
Keith J. Figlioli joined LRVHealth in 2016 as a Managing Partner after nearly two decades of experience building and growing healthcare IT and services companies. Before joining LRVHealth, Keith was Senior Vice President of Healthcare Informatics at Premier, Inc. (NASDAQ: PINC) where he led the company’s enterprise healthcare technology business unit and helped the company raise $820M in its IPO. Prior to Premier, Keith was Senior Vice President of Enterprise Solutions at Eclipsys (acquired by Allscripts NASDAQ: MDRX). He is a board member of IntelyCare, Season Health, Censinet, ArcheHealth and Reimagine Care, and a board observer for Diana Health and Genome Medical. Keith previously served on the Board of Trustees at Wheaton College, the Strategy Committee of Baystate Health, the Health IT Standards Committee for the Office of the National Coordinator (ONC) and the board of directors for Global Healthcare Exchange (GHX), Activate Networks, Good Sports, and MassBike. Keith earned his B.A. from Wheaton College, M.B.A. from Boston University, and M.P.H. from Harvard T.H. Chan School of Public Health.