What we’re seeing in the field has shifted. The conversation in healthcare AI has moved from “should we explore this?” to “how do we deploy this at scale without breaking something critical?” That is a meaningful change in posture, and it carries implications that most boards and clinical leadership teams are not yet equipped to handle.
This week brought two data points that, read together, paint a clear picture of where we are. NVIDIA’s 2026 healthcare AI survey confirmed that AI is delivering measurable return on investment across radiology, drug discovery, and clinical documentation — the ROI question is largely settled. And Novo Nordisk announced a full-enterprise partnership with OpenAI, integrating AI across drug discovery, clinical trials, manufacturing, supply chain, and commercial operations, with full deployment targeted by end of 2026.
On its own, each story sounds like a milestone. Together, they signal something more consequential: the healthcare sector is no longer running controlled pilots. It is moving AI into the operational core of clinical and commercial decision-making. And the governance infrastructure required to support that transition is, in most organizations, still in the planning stage.
The ROI Is There. The Readiness Largely Is Not.
NVIDIA’s survey data matters because it ends a debate that was consuming too much executive bandwidth. The question was never really whether AI could work in clinical environments — it was whether organizations could deploy it responsibly, at the right scale, with the right oversight. The ROI confirmation clears one hurdle and immediately raises a harder one.
Grant Thornton’s 2026 AI Impact Survey found that 78% of business executives lack strong confidence they could pass an independent AI governance audit within 90 days. In healthcare, where patient safety and regulatory exposure are embedded in every deployment decision, that number is not abstract. It is a liability profile. Three in four boards have approved major AI investments — but 48% have not set AI governance expectations and 46% have not integrated AI risk into ongoing oversight.
These organizations are not failing to act. They are acting — aggressively, in many cases. What they are missing is the governance architecture that makes that action defensible when something goes wrong.
What the Novo Nordisk Model Actually Tells Us
The Novo Nordisk-OpenAI partnership is worth examining specifically because of its scope. This is not a vendor agreement for a discrete tool. It is a full organizational transformation — AI woven into drug discovery, clinical research protocols, manufacturing processes, supply chain decisions, and commercial strategy simultaneously. Full deployment by end of 2026.
For a pharma company operating in highly regulated environments across multiple jurisdictions, that is an extraordinary governance commitment. It means model accountability frameworks, data provenance documentation, regulatory audit trails, and clinical validation protocols all have to be operational — not aspirational — before full rollout.
Most health systems and medtech companies watching this announcement will not have the internal infrastructure Novo Nordisk has built. That gap matters. The Novo Nordisk model is not something to copy directly. It is something to understand structurally — and then build toward at the pace your organization can actually sustain.
The Regulatory Floor Is Rising — and It Is Sector-Specific
On March 20, 2026, the White House released its National Policy Framework for Artificial Intelligence, recommending that Congress establish a minimally burdensome national standard and preempt state AI laws that impose undue burdens on innovation. The intent is to create regulatory clarity. The operational reality is that clarity has not arrived yet.
Since mid-March, 25 new state AI laws have been enacted. Another 27 bills have passed both legislative chambers. Federal preemption may eventually resolve this fragmentation — but Congress is not on a fast timeline, and in the meantime, healthcare organizations operating across state lines are navigating an accelerating patchwork of obligations.
What makes healthcare uniquely exposed is that sector-specific regulation has always existed alongside whatever general AI governance frameworks emerge. The FDA’s regulatory framework for AI-enabled medical devices, HIPAA obligations around data provenance, and clinical liability standards do not pause while Congress debates preemption. Health systems deploying AI at scale are subject to all of it simultaneously.
The Governance Question Has Changed
KPMG and INSEAD launched their Global AI Board Governance Principles this month — a recognition that governance frameworks designed for static enterprise software are inadequate for AI systems that learn, adapt, and increasingly operate with autonomous decision-making authority.
The governance question in agentic AI environments is no longer “what can this model generate?” It is “what can this system do on our behalf, and who is accountable when it acts?” In clinical settings, that question carries direct patient safety implications. A radiology AI that misses a finding, a clinical documentation tool that introduces error at scale, a supply chain optimization model that creates drug shortage — the accountability chain runs directly to the deploying organization.
Boards that approved the investment but did not build the governance structure are not off the hook. The regulatory environment is specifically moving toward demonstrable oversight — documented governance, not written policy. The difference matters enormously in a post-incident environment.
What This Means for Health System and MedTech Leaders Right Now
The practical implications for decision-makers navigating this environment:
Governance architecture is not the last step in an AI deployment — it is a precondition. Our AI governance advisory practice is built around this sequencing: building accountability infrastructure before deployment, not as a compliance exercise after. Organizations that build governance in parallel with pilots are far better positioned than those that treat it as a compliance exercise after deployment.
The board needs to be asking different questions. Approving AI spend is no longer sufficient. Board-level oversight now requires understanding what governance expectations are in place, how AI risk is monitored on an ongoing basis, and who is accountable when systems operate outside intended parameters.
Vendor evaluation must go deeper than the demo. In an environment where full-enterprise AI integration is the direction — not the exception — due diligence on AI vendors requires examining model accountability documentation, data governance practices, regulatory readiness, and what happens when the model is wrong. A strong demo tells you very little about any of this. For investors evaluating healthcare AI companies, our AI and XR investment advisory applies this same diligence framework to the capital decision.
The workforce integration question is underweighted. This governance gap extends well beyond healthcare — for the enterprise-wide picture across sectors, see our analysis of agentic AI and the governance infrastructure that isn’t keeping pace. AI tools that generate ROI in controlled conditions frequently encounter friction at scale because the clinical and operational workforce has not been integrated into the deployment process. Training is not sufficient. Change management, workflow redesign, and feedback loops are the difference between deployment and adoption.
The Strategic Posture That Will Hold
The organizations that will navigate this period with the least exposure are not the fastest movers or the most cautious. They are the ones that have made a deliberate choice: deploy AI where the value is demonstrable and the governance infrastructure can be built concurrently. Not AI everywhere. AI where it can be done well.
The ROI evidence is in. The regulatory pressure is accelerating. The board accountability standards are tightening. The organizations still treating governance as a future problem are running out of runway.
What we’re seeing in the field is that the margin for learning-while-governing has narrowed considerably. The organizations that recognized this 18 months ago are, right now, in a meaningfully stronger position than those that did not.
Reference Articles
Grant Thornton — 2026 AI Impact Survey Report
White House — National Policy Framework for Artificial Intelligence (March 20, 2026)
Holland & Knight — White House Releases a National Policy Framework for Artificial Intelligence
Alvarez & Marsal — The White House’s AI Legislative Framework and the Unsettled Future of State AI Laws
ACL Digital — Enterprise AI Healthcare in 2026: From Pilots to Scale
KPMG & INSEAD — Launch Global AI Board Governance Principles (April 2026)
WilmerHale — Board Oversight and Artificial Intelligence: Key Governance Priorities for 2026
Lynn Welch is Founder and Principal Advisor of The Lion’s View, an independent AI and extended reality advisory firm serving private equity, enterprise, healthcare, and defense organizations. She advises boards and executive teams on AI governance, vendor evaluation, and strategic risk — with no platform to sell and no vendor relationship to protect. Learn more about our advisory services.
Comments are closed