Who Owns the Algorithms? Inside Healthcare's Newest Leadership Role

By Matt Bell, DMSc, PA-C, CAQ-HM

If you believe your hospital doesn't really use artificial intelligence yet, let me describe what a clinical PA's typical day has become or soon will be. An algorithm scored their patients overnight for sepsis risk and decided which ones generated alerts. Another one read the chest imaging queue and flagged studies for the radiologist's early attention. A prediction model helped decide how many nurses were scheduled to your unit. Ambient software listened to a colleague's patient visit and drafted the note.  An insurer's model weighed in on whether your patient's procedure got authorized.

None of that is the distant future. All of it is becoming ordinary, and most of it is invisible, which sets up the question this article is actually about. When one of those tools quietly stops working as well as it used to, who notices? When two of them disagree, who decides? When a clinician asks, "Can I trust this alert?" whose phone rings? At most institutions, until very recently, the honest answer was that no one’s did. No single leader owned the algorithms.

If it is any comfort, the confusion is universal. In an April survey of boards and executives across industries, 90% of board members said the C-suite owns AI strategy, but the C-suite itself could not agree: barely a third said they collectively owned it, with the rest scattering responsibility among business-line leaders, functional heads, and management levels below.¹⁰ Corporate America, in other words, built the tools before deciding who answers for them. Healthcare simply has more at stake when the answer is nobody.

Healthcare is now starting to fix that, the way it has always fixed problems of unowned risk; by creating a leader and handing them accountability. The title emerging across the country is the Chief Health AI Officer, and even if you never hold the role, you are going to work with one. Here is what they actually do.

Why the role suddenly exists

As I said earlier, Medicine has a long habit of inventing new leadership when a new class of risk arrives. When hospital infections became undeniable, we created Infection Control Officers. When quality became measurable, the Chief Quality Officer followed. New risk, new steward, every time.

AI crossed that threshold quietly and all at once. The FDA's list of AI-enabled medical devices has surpassed 1,400 and continues to expand monthly.¹ The peer-reviewed literature began documenting the Chief Health AI Officer as an emerging executive role in 2024.² And in June, the Joint Commission launched its first-ever certification for the responsible use of AI, evaluating organizations across governance, data management, risk and bias reduction, monitoring and validation, and workforce education.⁴ Read those five domains again slowly: that is a job description. Someone in the building has to own the answers.

What the organizational charts now show

This is no longer theoretical, and the American Hospital Association has been documenting it. In a February analysis titled "The C-Suite Rewrites the Org Chart," the AHA described technology and AI leadership rising "to the same level of importance as clinical, financial and operational strategy" and named names.³ Cleveland Clinic has established a Chief Artificial Intelligence Officer accountable for enterprise AI strategy, governance, and safety to ensure tools are "clinically appropriate, ethically deployed and aligned with organizational goals." NYC Health + Hospitals, the nation's largest municipal system, formalized a Chief Data and AI Officer. Sanford Health, the largest rural health system in the country, consolidated technology, AI, analytics, and innovation under a single executive who is a physician, notably not a technologist.³

Two details in the AHA's account deserve a slow read. First, these roles increasingly report directly to the CEO.  This is a reporting line that, in the AHA's words, "underscores the strategic importance of the position." Second, boards themselves are now "seeking assurance that digital investments support organizational strategy and that emerging technologies are deployed responsibly." When board governance starts asking a question, an executive eventually has to own the answer. The AHA's summary of where the field has landed is blunt: the debate is no longer "whether technology should have a seat at the table" but "what kind of seat and how much authority it requires."³ There is, the AHA concedes, "no single blueprint," which means the institutions building these roles now are the ones writing it.

What the job actually is

Strip away the futuristic title and the work sorts into six very recognizable kinds of leadership.

Managing the AI Formulary (Which Algorithms Are Approved, Used, and Replaced). No hospital adds a new medication to its formulary because a sales rep was persuasive; a pharmacy and therapeutics committee weighs the evidence, approves it, and tracks it. Until now, algorithms had no equivalent as these tools arrived through a purchasing decision, a pilot, or an EHR update, and simply began touching patients. The AI officer's first job is building that missing structure: an oversight committee with clinical voices, subcommittees that evaluate tools before deployment, and a living inventory of what is actually in use. Less science fiction, more air-traffic control.

Evaluating the tools (and owning the risk). Before any tool reaches that inventory, someone must hold the authority to look a vendor in the eye and ask for proof: evidence the tool works, evidence it works for patients like ours, and an honest accounting of what happens when it fails. That evaluative power and the risk management that follows from it sit at the center of this role. It means naming each tool's failure modes before deployment, deciding on mitigations (what stays under human review, what the fallback is when the system is down or wrong, how incidents are reported and learned from), and revisiting that ledger for as long as the tool is in use.⁴ However, here is what makes healthcare's version unlike any other industry's: the risk ledger runs three layers deep. The health system carries the financial and legal exposure. The staff inherits the tool's mistakes as their own workload and liability. But the deepest line belongs to the patient who never chose the algorithm, and usually never knows it was in the room.

Watching for drift. Here is the property of AI that surprises clinicians most: an FDA-approved drug has the same molecular structure in year five as on approval day, but an algorithm's performance decays as patients, practice patterns, and documentation habits change around it. A sepsis model tuned to 2024's patients slowly becomes wrong about 2027's patients, and this happens without an error message or a recall notice. Someone must run surveillance on AI software the way we run it on infections. Today, surveys suggest only about one in ten healthcare organizations has automated monitoring of its AI.⁵ That is the gap this role exists to close.

Leading the humans, not the machines. The dirty secret of health technology is that it rarely fails technically; it fails socially. Consider the most well-known AI tool in medicine right now, the ambient scribe: a 2026 multisite study found it saved clinicians 13-16 minutes per day, yet only about a third used it consistently.⁶ The distance between a tool that works and a tool that is used is not an engineering problem. It is trust, workflow, and change leadership, which is why the AI officer spends far more time with people than with servers.

Teaching the workforce. Across the professions, only about 14% of clinicians say their training prepared them to work with AI⁷ while a majority are already using it, self-taught and unsupervised.⁸ An organization cannot govern what its people don't truly understand. Building that literacy, from the boardroom to the front lines, sits squarely in this portfolio.

Translating. Above all, the role is a translation post. The data scientists, executives, and clinicians in an organization do not speak the same language, and every AI decision requires all three groups to understand one another. The AI officer is the person who can stand in each one and be believed.

The competing demands

Listed that way, the job sounds orderly: six duties, a tidy stack of committees, done. What the list hides is that every one of those duties is performed under competing demands, and those demands often work against each other. Four of them define the role’s real difficulty.

First, this leader must govern a technology the workforce is already using. When a P&T committee evaluates a new drug, the drug waits outside until it's approved. AI never waited: a majority of clinicians are already using it in some form, self-taught and largely unsupervised.⁸ Clamp down too hard, and the use simply goes underground, beyond anyone's sight; move too permissively and unvalidated tools keep touching patients. The AI officer is building the airport while the planes are already landing.

Second, the rulebook is still being written. The FDA regulates AI only when it lives inside a medical device, but much of the AI in a hospital, from ambient scribes to scheduling models to the chatbot in the patient portal, sits partly or wholly outside device regulation. Also, the Joint Commission certification described earlier (the closest thing to a rulebook that exists) is one organizations can choose to pursue; nothing requires it. So, there is no compliance manual to hide behind. This leader must set internal standards in advance of external requirements, then defend those standards to people who would rather wait to be told.

Third, every constituency pulls in a different direction, and each one is right. The CFO wants the efficiency the vendor promised. The CIO wants clean integration. Clinicians want their time back but not another alert. Compliance wants zero risk in a domain where zero risk means zero use. Patients, when asked, want to be told when AI touched their care. There is no decision that satisfies that whole room; the job is adjudicating among legitimate interests, in public, repeatedly, which is why it wears down leaders who mistake it for a technical assignment.

And fourth, the role is designed to lead across boundaries rather than to absorb everyone else’s portfolios. The sepsis model sits in quality’s workflow; the ambient scribe cuts across clinical operations, IT, and revenue cycle; the imaging tool sits in radiology; and the data sits with IT. This leader may have direct reports, but many of the outcomes they are responsible for are delivered through peers’ departments rather than their own. Their leverage is less about hierarchy and more about influence: coalitions, credibility, and change leadership alongside other executives. Add the pace problem, as governance committees meet monthly while the underlying models update continuously, and you begin to see why this is less a new job than a new discipline.

It is fair to note that some management thinkers, writing in Harvard Business Review, have argued the opposite of this article's premise, saying that AI is too broad for any single leader and belongs to a distributed team instead.¹¹ Healthcare's emerging model quietly absorbs that critique rather than contradicting it. Look again at the anatomy above: the execution is already distributed among committees, subcommittees, department owners, and frontline champions. What is not distributed is the accountability. Medicine learned that distinction the hard way, through every quality and safety failure that had many parents and no owner. Distributed execution, concentrated accountability: that is the design, and it is why the seat exists.

Notice what's missing

Now look back across the duties and the competing demands and notice what appears nowhere: writing code. Evaluating evidence, chartering committees, weighing and mitigating risk, surveilling performance, leading change, educating a workforce, adjudicating among rivals with legitimate claims, and you see this is a leadership job wearing a technical title. That is precisely why organizations are increasingly filling it from the clinical ranks with physicians, nurses, and PAs who have spent their careers earning the one currency this role runs on: trust. An algorithm can be brilliant, but it cannot be trusted by a night-shift nurse who has never met it. A clinician-leader can carry that trust into the room on its behalf and, just as importantly, can refuse to endorse it when the evidence is not there.

The doctoral PA's stake

Which brings this home to this blog's readership. Lay this role's portfolio side by side with a doctoral PA's background, and the overlap is almost uncomfortable. Evidence appraisal is the spine of the DMSc-, DHSc-, DrPH-, EdD-, DHA-, and DPA- prepared PA. Quality improvement and systems thinking (the disciplines behind surveillance and monitoring) are second nature, as are change management and organizational leadership. Education is in the profession's bones, in a field where precepting the next generation is simply assumed. The AHA says there is "no single blueprint" for who holds these roles.³ But draft one from the competencies alone, and it would read like a doctoral PA's résumé.

There is a deeper resonance here, and doctoral PAs will feel it. Ours is a profession that exists because medicine once faced a workforce gap it could not close with physicians alone and answered by creating and trusting a new kind of clinician. Over the decades, PAs have moved from an experiment at the bedside into essential partners in care delivery, education, and leadership. Sixty years later, healthcare faces a similar type of problem, and the founders’ model still applies: do not wait for permission; demonstrate capability where it is needed. This time the need runs through both worlds doctoral PAs inhabit: the health systems deploying these tools and the academic institutions working out how to teach, govern, and study them, where the same portfolio of governance, evidence, education, and change leadership applies. The seat is new. The preparation, for this readership, is not.

The question to take to work

You do not need to want this job to take something from this article. Consider this question instead: who owns the algorithms where you work? Ask it at your next staff meeting or leadership huddle. If the answer is a name, you have learned your organization is ahead of most. If the answer is silence, you have learned something more important and, with 87% of healthcare executives admitting their organizations lack adequate AI guidance, you will not be the only one learning it.9

Healthcare has already decided that the algorithms need an owner. The next few years will decide who those owners are and whether PA leaders, who have already earned trust in patient care, education, and leadership, choose to step forward.

  

References

1. US Food and Drug Administration. Artificial intelligence-enabled medical devices. FDA.gov. Accessed July 11, 2026. https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-enabled-medical-devices 

2. Beecy AN, et al. The chief health AI officer — an emerging role for an emerging technology. NEJM AI. 2024. doi:10.1056/AIp2400109

3. American Hospital Association, Center for Health Innovation. The C-suite rewrites the org chart: how health systems are elevating technology leadership. AHA Market Scan. February 3, 2026. Accessed July 10, 2026. https://www.aha.org/aha-center-health-innovation-market-scan/2026-02-03-c-suite-rewrites-org-chart-how-health-systems-are-elevating-technology 

4. The Joint Commission. Joint Commission releases first of its kind, exclusively designed for healthcare organizations, voluntary Responsible Use of AI in Healthcare certification. 2026. Accessed July 14, 2026. https://www.jointcommission.org/en-us/knowledge-library/news/2026-05-responsible-use-of-ai-in-healthcare-certification 

5. Censinet, College of Healthcare Information Management Executives (CHIME). AI adoption survey reveals healthcare's governance gap and drive toward agentic usage. 2025. Accessed July 11, 2026. https://chimecentral.org/chime/resource-press-release/ai-adoption-survey-reveals-healthcares-governance-gap-drive-toward-agentic-usage 

6. Rotenstein LS, Holmgren AJ, Thombley R, et al. Changes in clinician time expenditure and visit quantity with adoption of artificial intelligence–powered scribes: a multisite study. JAMA. 2026;335(16):1408-1417. doi:10.1001/jama.2026.2253

7. Chatzichristos C, Chatzichristos G, Borremans I, et al. Bridging the AI-literacy gap in health care: qualitative analysis of the Flanders case study. J Med Internet Res. 2025;27:e76709. doi:10.2196/76709

8. DeStefano, J. Wolters Kluwer survey: physician assistant workforce steps up amid clinician shortages, fragmented care landscape, and changing patient demands. 2025. Accessed July 10, 2026. https://www.wolterskluwer.com/en/news/wolters-kluwer-survey 

9. Digital Medicine Society (DiMe). 3 key insights for the 2026 health AI horizon. January 2026. Accessed July 12, 2026. https://dimesociety.org/newsroom/blog/3-key-insights-for-the-2026-health-ai-horizon/ 

10. Boards say the C-suite owns AI strategy. The C-suite doesn't agree. Fortune. April 22, 2026. Accessed July 10, 2026. https://fortune.com/2026/04/22/ai-ownership-c-suite-board-disagree-pearl-meyer-survey-brad-jayne/ 

11.Winsor J, Stave J, Kurt R. Your AI strategy needs more than a single leader. Harvard Business Review. August 4, 2025. Accessed July 10, 2026. https://hbr.org/2025/08/your-ai-strategy-needs-more-than-a-single-leader

Matt Bell, DMSc, PA-C, CAQ-HM is the Chief AI Officer and Chief PA of the Western North Carolina VA Health Care System, a process improvement medical officer, and a PA educator with 30 years of clinical experience and 28 years precepting PA students. He writes and speaks nationally on AI governance, implementation, and leadership in healthcare. The views expressed are his own.