Improvement or Transformation: How Boards Should Classify AI Proposals

Every executive knows the difference between incremental improvement and structural transformation. Applied to artificial intelligence, the same distinction decides what the board approves, what conditions it attaches, and what the minutes must show.

In Brief

The Problem. Artificial intelligence (AI) proposals reach health system and life sciences boards in the same form whether they refine a process or change who makes a clinical, coverage, or regulatory decision.

Why It Persists. Management frames most AI as efficiency, vendors add AI features inside routine software updates, and boards have no agreed test for telling an improvement from a transformation.

What Boards Should Do. Require management to classify every AI proposal before it reaches the board, test that classification with four questions, govern improvements through the inventory and the framework, and approve transformations as decisions with conditions attached.

Leaders in healthcare and life sciences drive growth through two approaches: business improvement and business transformation. They are complementary but distinct. Improvement brings incremental, evolutionary advances. Transformation brings disruptive or revolutionary change. When I first wrote about this distinction in 2022, I treated it as a question of strategy. My argument now is that it has become a question of governance. AI proposals fall on one side of the line or the other, and the oversight each requires is different. A board that treats every AI proposal the same way will either slow routine improvements or approve transformations with too little scrutiny.

Business Improvement

Business improvement builds on existing systems and processes to raise performance. Its gains are incremental, and they compound over time. Personalized medicine shows how. Researchers have improved genomic data analysis step by step, and its integration with patient medical histories, to deliver more personalized treatment. No single step transformed the field; together they changed what is possible.

Three practices mark successful improvement:

• Incremental changes. Small, continuous improvements that yield significant results over time.
• Focus on efficiency. Streamlined processes that reduce waste and make better use of resources.
• Performance metrics. Data and analytics that identify opportunities and measure results.

AI as Improvement

An AI improvement makes an existing process faster, cheaper, or more consistent. It does not change who decides or what is decided. Ambient documentation is the best-evidenced example. In a randomized trial across ambulatory clinics in two states, ambient AI scribes reduced practitioners' work exhaustion and cut time spent on notes by about 0.36 hours a day, without compromising note quality or billing compliance (Afshar et al., 2025). The clinician still reviews and signs the note. The accountable person and the decision stay the same.

For these proposals, the board's role is to confirm that each system operates inside a governed framework, not to approve each one. The National Institute of Standards and Technology (NIST) places this work in the "Govern" function of its AI Risk Management Framework, which covers policies, roles, and accountability across all systems (National Institute of Standards and Technology, 2023).

Improvement does not mean low risk. A documentation tool handles protected health information. An AI feature can arrive inside a routine vendor software update, with no proposal at all. The board should therefore expect every improvement to appear in the organization's AI inventory, under a named accountable executive, with performance reported on a schedule.

Business Transformation

Business transformation changes the organization's core structure, processes, culture, or business model. It aims at disruptive or revolutionary gains and alters the organization's trajectory and market position. In healthcare and life sciences, I see it most often in a move to value-based care, in investment in new research capability, and in partnerships that change how the organization creates value.

Three practices mark successful transformation:

• Comprehensive change. Coordinated changes that reach many parts of the organization at once.
• Strategic vision. Initiatives aligned with long-term organizational goals.
• Cultural shift. A mindset of adaptability and innovation across the organization.

AI as Transformation

An AI transformation changes who makes a decision, what evidence supports it, or how the organization earns revenue. Examples include a model that determines prior-authorization outcomes, a clinical decision support tool that changes care pathways across a system, and AI-generated real-world evidence offered in a regulatory submission. These proposals change the organization's risk profile. Some change its relationship with regulators, patients, and courts.

Litigation shows how. In Estate of Lokken v. UnitedHealth Group, plaintiffs allege that the insurer used an AI tool to replace physicians' judgment in post-acute coverage decisions for Medicare Advantage members; the company disputes the allegations and says physicians make the determinations. On March 9, 2026, a federal magistrate judge ordered discovery that reaches the insurer's internal AI review board and the compensation of medical directors (Greenberg et al., 2026). Whatever the outcome, the case shows that the governance record behind a decision-making system will be read by others.

For life sciences companies, the regulator's own terms draw the line. In January 2025, the Food and Drug Administration (FDA) issued draft guidance proposing a risk-based framework for establishing the credibility of an AI model for a particular context of use (Food and Drug Administration, 2025). As of October 21, 2026, that guidance remains in draft. A board that approves AI-generated evidence as a strategic initiative should ask first how management will establish its credibility. Time saved is a secondary question.

Transformations belong before the full board or its designated committee as decisions, with conditions attached. In Marchand v. Barnhill, 212 A.3d 805 (Del. 2019), the Delaware Supreme Court held that directors must make a good-faith effort to establish board-level monitoring of mission-critical compliance risks (Duffy et al., 2019). AI that decides coverage, shapes clinical care, or supports regulatory submissions meets that description in most healthcare and life sciences organizations. The standard is procedural. The minutes are the evidence that the board met it.

What Both Require

Despite their differences in scope and scale, improvement and transformation share five requirements:

• Clear goals. Both aim to strengthen the organization's performance, competitiveness, and sustainability.
• Focus on patients and customers. Initiatives succeed when they deliver better value, services, or products to the people the organization serves.
• Change management. Both depend on engaging stakeholders, aligning the organization, and implementing new processes, technologies, or structures with discipline.
• Decisions grounded in data. Data and analytics identify opportunities, measure progress, and inform decisions.
• Continuous learning. A culture of learning and adaptability sustains both kinds of change.

With AI, the last two carry a governance meaning. The data that measure progress must include the system's own performance, and learning must include what the organization does when a system underperforms.

The Classification Test

Management should classify each AI proposal before it reaches the board, and the board should test that classification. Four questions are usually enough:

• Does the system make, or materially inform, a decision that a named person made before?
• Does it affect patients, members, or regulatory submissions directly, or only internal operations?
• Would its failure be reportable to a regulator, an accreditor, or a payer?
• Can the organization switch it off without stopping a core process?

A "yes" to any of the first three questions, or a "no" to the fourth, signals transformation. The test is conservative by design. In my experience, the failure that surfaces later, in an incident review or in litigation, is rarely the transformation the board debated. It is the one that arrived labeled as an improvement.

Running Both Streams at Once

Most organizations need both kinds of change at the same time. Two illustrations show how they combine.

Healthcare: integrated care delivery. A hospital system sets out to improve care coordination across its facilities. As an improvement, it implements an electronic health record (EHR) system that streamlines clinical documentation and communication among clinicians. As a transformation, it restructures around integrated care teams of primary care physicians, specialists, nurses, and allied health professionals who manage patients across the whole continuum of care. Today the same system might add ambient documentation as an improvement while it redesigns those teams around AI-supported triage as a transformation.

Life sciences: digital transformation. A pharmaceutical manufacturer sets out to use digital technology and advanced analytics to drive innovation and efficiency. As an improvement, it adopts cloud platforms for data storage, collaboration, and real-time access for researchers and decision makers. As a transformation, it invests in real-world data (RWD), AI, and machine learning to generate real-world evidence (RWE) on the safety of new molecular entities (NMEs) in real-world populations, with the aim of supporting regulatory decisions and reducing development cost and time. The second stream is the one the FDA's draft credibility framework addresses.

The board's task is to keep the two streams separate in its reporting. Otherwise a long list of modest improvements can hide the one transformation traveling with them.

Opportunities and Challenges Ahead

The same distinction helps a board read the developments that will shape the industry over the next several years:

• Digital health. Telemedicine, remote patient monitoring, and digital therapeutics can make care more accessible and less costly. Many deployments are improvements to existing services; those that change who triages or treats a patient are transformations.
• Precision medicine. Genomic, proteomic, and other molecular data support targeted therapies and more effective treatment plans. AI that selects patients or therapies falls on the transformation side of the line.
• Value-based care. Payment tied to outcomes and cost requires organizations to reconsider strategy and operations. Moving a population into risk-based contracts is a transformation, and the analytics that support it should be governed as one.
• Regulatory environment. The Joint Commission launched its voluntary Responsible Use of AI in Healthcare certification on June 1, 2026, with standards that include governance, monitoring of safety performance, and bias reduction (Guerra, 2026). Colorado's replacement AI law, S.B. 26-189, takes effect on January 1, 2027. It requires notice before automated decision-making technology materially influences a consequential decision, and its exemption for entities covered by the Health Insurance Portability and Accountability Act (HIPAA) is partial: they still owe patients a general notice (Darch et al., 2026). These rules reach the transformation category first, and a board that already classifies its AI will know which deployments they touch.
• Workforce development. A skilled and adaptable workforce requires continuous learning. Half of the healthcare organizations in a September 2026 survey of 226 executives had cut headcount because of AI or planned to within six months (Guerra & Kraus, 2026). A reduction of that kind is a transformation, whatever the proposal calls it.
• Partnerships and collaboration. Partnerships with other organizations, academic institutions, government agencies, and patient advocacy groups share knowledge, resources, and expertise. For AI, the partnership terms (rights to data, change notification, audit, and liability) decide much of the risk.

The Strongest Objection: The Line Is Too Blurry

The strongest objection is that the line between improvement and transformation is too blurry to govern by. A documentation tool can drift into clinical suggestions; a scheduling optimizer can begin to decide which patients are seen first. Classification, the objection runs, becomes paperwork that slows useful tools.

The drift is real, and it is the reason to classify, not a reason to stop. The test asks about the decision and the consequence, not the technology, so a system that drifts changes category and the inventory shows it. The cost is small: most proposals are improvements and pass through the framework without a board vote. What the board gains is the ability to show that it saw the few transformations and decided them deliberately.

Organizations that combine evolutionary and revolutionary change, and govern each for what it is, will be better placed to serve their stakeholders and, above all, their patients.

For the Board

• Has management classified each AI system in the inventory as an improvement or a transformation, and who reviewed that classification?
• Which AI transformations has the board or a committee approved as decisions, and what conditions did it attach?
• How does the board learn of AI features that vendors add to existing products, and of systems whose use has drifted since approval?
• For AI-generated evidence in a regulatory submission or an in-licensed asset, how has management established the model's credibility for its context of use?
• Which of our AI deployments would Colorado's notice rules or the Joint Commission certification standards reach, and are our records ready for that review?

Find the board questions for your committee in the Avancer AI Governance Navigator.

References

Afshar, M., Baumann, M. R., Resnik, F., Hintzke, J., Sullivan, A. G., Wills, G., Lemmon, K., Dambach, J., Ann Mrotek, L., Quinn, M., Abramson, K., Kleinschmidt, P., Brazelton, T. B., Leaf, M. A., Twedt, H., Kunstman, D., Patterson, B., Liao, F., Rasmussen, S., . . . Gordon, J. (2025). A pragmatic randomized controlled trial of ambient artificial intelligence to improve health practitioner well-being. NEJM AI, 2(12). https://doi.org/10.1056/aioa2500945

Darch, J. E., Brooks, P., Daudi, S., Hinckley, D., & Kaye, A. (2026, May 19). Colorado scales back AI law, with targeted implications for health care. Ropes and Gray. https://www.ropesgray.com/en/insights/alerts/2026/05/colorado-scales-back-ai-law-with-targeted-implications-for-health-care

Duffy, M. P., Ritts, G. J., & Heintz, A. F. (2019, August). Delaware Supreme Court reinforces directors' oversight obligations on mission-critical subjects. Jones Day. https://www.jonesday.com/en/insights/2019/08/delaware-court-reinforces-directors-oversight

Food and Drug Administration. (2025, January 7). Considerations for the use of artificial intelligence to support regulatory decision-making for drug and biological products [Draft guidance]. U.S. Department of Health and Human Services. Retrieved October 21, 2026, from https://www.fda.gov/regulatory-information/search-fda-guidance-documents/considerations-use-artificial-intelligence-support-regulatory-decision-making-drug-and-biological

Greenberg, D. S., Heilman, K., & Shlaferman, H. Z. (2026, April 23). Federal court orders broad discovery against UHC in AI coverage denial lawsuit. ArentFox Schiff. https://www.afslaw.com/perspectives/alerts/federal-court-orders-broad-discovery-against-uhc-ai-coverage-denial-lawsuit

Guerra, A. (2026, June 1). Joint Commission launches AI certification as adoption outpaces oversight. Health System CIO. https://healthsystemcio.com/2026/06/01/joint-commission-ai-certification/

Guerra, S., & Kraus, S. (2026, September 30). The 2026 healthcare AI ROI scorecard. Bessemer Venture Partners. https://www.bvp.com/atlas/the-2026-healthcare-ai-roi-scorecard

National Institute of Standards and Technology. (2023). Artificial intelligence risk management framework (AI RMF 1.0) (NIST AI 100-1). U.S. Department of Commerce. Retrieved October 21, 2026, from https://www.nist.gov/itl/ai-risk-management-framework