What Six Corporate Failures Teach Boards About Overseeing AI

Updated October 21, 2026. First published February 21, 2024.

Kodak, Blockbuster, Nokia, Volkswagen, Wells Fargo, and Uber are usually told as stories about technology. Read against the public record, they are stories about information that reached the board too late, or not at all, and that is the lesson health and life sciences boards need for artificial intelligence.

In Brief

The Problem. Boards are approving and inheriting artificial intelligence (AI) faster than they can see it. In PwC's 2026 survey of about 600 public company directors, 60 percent said AI is not yet part of board oversight, and only 18 percent rated the metrics linking AI outcomes, risks, and performance as good or excellent (Gaskin, 2026).

Why It Persists. The familiar case studies teach the wrong lesson. They tell directors to foresee the next technology. The record shows that most of these companies had the technology, the data, or the warning signs. What they lacked was a designed path for that information to reach the board, and a board habit of asking for it.

What Boards Should Do. Treat AI oversight as a reporting system the board designs, not a forecast it makes. Require an inventory of AI in use, a named accountable executive, a committee charter that assigns oversight, conditions on every material approval, and minutes that show what the board asked and decided.

Learning From Past Technological Adaptation

The pace of technological change is reshaping industries and business models, and it presents corporate boards with both challenges and opportunities. Business schools teach the best-known corporate failures of the past 25 years as warnings about disruption: companies that failed to anticipate a technology or a shift in the market. Boards, the usual telling goes, must balance innovation against stability and compliance, and the forward-looking ones will prevail.

When I rebuilt these cases from primary sources, a different pattern emerged. Kodak invented the digital camera. Nokia led the smartphone market. Blockbuster was offered Netflix. Wells Fargo's sales practices were reported in the press nearly three years before regulators fined the bank. The failures were rarely failures of technology or even of foresight. They were failures of what the board saw, what it asked, and what it recorded.

That distinction is the argument of this article, and it matters for health systems, payers, and life sciences companies now deploying AI. The board does not run the models. Its duty is to know where AI is used, to understand the material risks, and to show in the minutes that it asked. I describe this posture as "nose in, fingers out." The rest of this article tests that argument against six failures, six opportunities, and eight areas where the board's role is expanding.

Learning From Experience

Since 2001, Avancer has advised boards, chief executives, and their organizations in healthcare, life sciences, medical technology, real-world data and evidence, and AI. In that work, as an advisor and as a chief executive, I have seen technological change succeed or fail at three levels, each with a distinct job.

• The board as overseer of change. The board sets direction, approves material commitments with conditions, and holds management to account. It asks the questions that test a strategy; it does not write the strategy.
• The chief executive as leader of change. The chief executive turns direction into a plan, builds the team, and owns the result, including the duty to tell the board promptly when something goes wrong.
• The organization as practitioner of change. Managers and staff adopt, validate, and run the tools. Their experience is the earliest warning the board will receive, if the reporting path carries it upward.

The cases that follow show what happens when one of those levels does another's job, or when the path between them breaks.

AI: Expanding Board Challenges

Boards face a widening set of challenges from AI: complex technologies, new categories of risk, and a growing body of regulation. The six cases below are the ones the original version of this article used. I have corrected each against the public record and drawn the lesson for a health or life sciences board.

Standing in the Past: Kodak

A Kodak engineer, Steven Sasson, built the first digital camera in 1975. Executives saw no market for it, and the device never reached production (Goodrich, 2022). Kodak did not ignore digital photography, however. By 2005 it led digital camera sales in the United States, with a 24.9 percent share (Associated Press, 2006). It still filed for Chapter 11 protection on January 19, 2012, with USD 6.75 billion in debts (CNNMoney, 2012).

The usual telling says Kodak's board refused to embrace digital. The record says something harder. Kodak embraced digital and won market share, but the new product line never replaced the economics of film. What Kodak missed, in one analysis of the case, was the business model, not the technology (Anthony, 2016). No public source I could find attributes the decisions to the board rather than to management, so the lesson is not about one board's stubbornness. It is about a board that measured adoption rather than economics.

Board lesson. Health and life sciences boards face the same trap with AI. A program can meet every adoption target while the margin of the service line keeps falling. Directors need enough grasp of where AI creates value, and where it does not, to ask the right question. The evidence is uneven. In a September 2026 survey of 226 healthcare executives, self-reported returns averaged about 3.5 times spend, led by provider revenue cycle work at 4.0 times, while provider clinical uses returned 2.9 times (Guerra & Kraus, 2026). Those figures are self-reported, which is itself a reason for the board to ask how management measures them. The question for directors is how each material AI investment changes the economics of the business, not only whether it went live.

Navigating Through Shifting Circumstances: Blockbuster

In 2000, the founders of Netflix offered their company to Blockbuster for USD 50 million. Chief executive John Antioco declined. Netflix was then an unprofitable rental business that mailed discs to subscribers (Mollman, 2023). It did not offer streaming until January 16, 2007 (Netflix, 2007). Blockbuster filed for Chapter 11 protection on September 23, 2010 (Ricker, 2010).

The familiar version blames Blockbuster's board for missing streaming in 2000. The record shows a management decision about a mail-order competitor, seven years before Netflix streamed a film. The board did shape what followed. In Antioco's own account, the activist investor Carl Icahn won a proxy fight and board seats for himself and two others, which left Antioco defending a growth strategy that included an online business and the end of late fees. Six months after a boardroom dispute over his bonus, Antioco left, and Blockbuster filed for bankruptcy three years later (Antioco, 2011). The lesson for boards is about visibility, and about whether a divided board can hold to a transition that costs money before it earns any. Which strategic decisions reach the board at all, and does the organization have the people and culture to act when the market moves?

Board lesson. In health care, many AI decisions are made inside operations, and many are made by staff without approval. In a September 2026 survey, 72 percent of healthcare leaders said AI tools are deployed without information technology approval at least occasionally, and more than a quarter had already implemented agentic AI (Hughes, 2026). Use among clinicians has moved fast as well. In 2026, 81 percent of physicians reported using AI, up from 38 percent in 2023, and 85 percent wanted to be consulted on adoption decisions (American Medical Association, 2026). The board should define in writing which AI decisions require its approval, and on what conditions. It should also ask management for a workforce plan: who will be trained, who will be redeployed, and how the organization will recruit the expertise it lacks.

Minimizing Potential Pitfalls: Volkswagen

Volkswagen fitted about 11 million diesel vehicles worldwide with software that defeated emissions tests (Washington Times, 2017). On January 11, 2017, it agreed to plead guilty in the United States and pay USD 4.3 billion in criminal and civil penalties. The United States case covered about 590,000 vehicles (U.S. Environmental Protection Agency, 2017). By March 2020, Volkswagen reported total costs of EUR 31.3 billion (Reuters, 2020).

The case shows that code can carry legal and ethical risk that no one sees unless someone asks how it behaves under test. The financial and reputational cost was severe. What the public record does not establish is the claim, in the original version of this article, that the board's lack of oversight caused the misconduct. The record does show what Volkswagen's supervisory board concluded afterward. A law firm it engaged found that former chief executive Martin Winterkorn breached his duty of care by failing, from July 27, 2015, to clarify promptly and comprehensively the use of unlawful software functions, and Volkswagen sought damages from him and from a former chief executive of its Audi unit (Associated Press, 2021). The failure the supervisory board named was one of escalation. The lesson I draw is narrower and more useful: software can be validated in one setting and behave differently in another.

Board lesson. AI raises the same question in a sharper form, together with risks to security, privacy, fairness, and accountability. A clinical algorithm can perform well in validation and differently in production, and vendor claims about performance deserve the same scrutiny. In September 2024, the Texas Attorney General settled with Pieces Technologies, whose generative AI tools were used in at least four Texas hospitals and which had advertised a "severe hallucination rate" of less than one per 100,000. The company denied wrongdoing and agreed, for five years, to disclose what such metrics mean and how they are calculated, or to retain an independent auditor (Fox, 2024). Regulators have begun to require the monitoring that follows. The federal rule under Section 1557 of the Affordable Care Act required covered entities to review patient care decision support tools for potential bias by May 1, 2025, although enforcement under the current administration is uncertain (Pressly & Leeds, 2025). California's SB 503, effective January 1, 2027, requires developers and deploying health facilities to identify and reduce the risk of biased results from clinical decision support tools, and requires deployers to monitor them (Montano & Tran, 2026). Directors should ask whether each high-risk AI system is monitored after deployment, by whom, how often it is audited, and how the results reach the board.

Optimizing the Allocation of Resources: Nokia

Nokia built an internet-ready touchscreen phone prototype in 2004 and then cancelled it. Reports attributed the decision to the cost of the touchscreen and the perceived risk (Hardawar, 2010). In 2010, Nokia still shipped about 100 million smartphones, more than twice Apple's volume. On February 11, 2011, it announced that Microsoft's Windows Phone would replace its own Symbian software as its primary platform (Lev-Ram, 2011).

Nokia did not miss the smartphone; it led the category. It missed the shift from devices to platforms, where Apple's iOS and Google's Android won developers and users. Its continued investment in Symbian, and the timing of its move to a partner's platform, were resource allocation decisions. By the time it partnered with Microsoft, its competitors controlled the ecosystems.

Board lesson. AI investment requires substantial resources, and boards must weigh it against current systems and core operations. The evidence on returns is mixed. Healthcare buyers in the 2026 survey cited above reported payback in about 12 months against the 24 months they expected (Guerra & Kraus, 2026). Yet only about one-third of surveyed health systems and health plans operate AI at scale (Janisch et al., 2025), and 59 percent of public company directors rate the information they receive on AI investments and returns as fair or poor (Gaskin, 2026). Partnerships with technology providers, research institutions, and other organizations can spread cost and risk. They also create Nokia's platform question. When the organization licenses a vendor's model, who controls the data, the updates, and the exit? The board should set minimum vendor terms (no training on the organization's data, notice of material model changes, audit rights, and indemnification) and ask whether each material contract meets them. Management negotiates the contracts.

Ensuring Adherence to Regulations: Wells Fargo

The Los Angeles Times described aggressive sales practices at Wells Fargo in December 2013, including accounts opened to meet quotas (Chittum, 2014). On September 8, 2016, the Consumer Financial Protection Bureau (CFPB) fined the bank USD 100 million. Penalties to the Office of the Comptroller of the Currency and to Los Angeles brought the total to USD 185 million (Consumer Financial Protection Bureau, 2016). The estimate of potentially unauthorized accounts later rose to about 3.5 million (Bloomberg, 2017).

Chief executive John Stumpf resigned in October 2016. The board released its own investigation on April 10, 2017. It found that the decentralized structure let business leaders resist oversight and understate problems in their reports to the board, and the board terminated senior managers and took compensation actions (Wells Fargo & Company, 2017). Commentators drew a direct lesson: boards need reports from control functions, not only from business leaders, and regular data on complaints and whistleblower matters (Kohn & Marcogliese, 2017). The former head of retail banking pleaded guilty in March 2023 to obstructing a bank examination (Schroeder, 2023). The Federal Reserve lifted its 2018 asset-growth restriction only on June 3, 2025 (Board of Governors of the Federal Reserve System, 2025).

Of the six cases, this one bears most directly on AI oversight. Delaware law treats oversight of mission-critical risk as a core director duty (In re Caremark International Inc. Derivative Litigation, 1996). In 2019, the Delaware Supreme Court sustained such a claim against the directors of an ice cream maker. The company had no board-level system for monitoring food safety, which the court called "mission critical" (Marchand v. Barnhill, 2019). On September 2, 2025, the Court of Chancery let a similar claim proceed against directors of the pharmaceutical company Teligent. The complaint alleged no board committee for compliance with Food and Drug Administration (FDA) requirements and a framework that left management to decide whether and when to report regulatory issues to the board (Bonadurer, 2026).

Board lesson. Regulation of AI is changing faster than any compliance calendar. Privacy laws still apply, including the European Union's General Data Protection Regulation (GDPR) (Regulation (EU) 2016/679, 2016) and the California Consumer Privacy Act (CCPA), as amended by voters in 2020 (State of California Department of Justice, n.d.). AI-specific law has now joined them. As of July 1, 2026, 29 states had enacted 109 AI laws that year, including at least six restricting how health insurers use AI (Brennen, 2026). Colorado replaced its 2024 AI Act with Senate Bill 26-189, signed May 14, 2026, and effective January 1, 2027. The new law requires notice before automated tools influence consequential decisions and a disclosure after adverse outcomes, with a partial exemption for entities covered by the Health Insurance Portability and Accountability Act (HIPAA), which must still give patients a general notice (Darch et al., 2026). In Europe, Regulation (EU) 2026/1744 entered into force on July 27, 2026, and moved high-risk obligations to December 2, 2027 for standalone systems and August 2, 2028 for AI embedded in regulated products (Hunton Andrews Kurth, 2026). Federal preemption remains unresolved: a December 11, 2025 executive order created a Department of Justice task force to challenge state AI laws (Ghavi et al., 2025). The board's job is not to track every rule. It is to ask which rules apply to its deployments and when, and to hear the answer from compliance as well as from the business.

Considerations of Ethics: Uber

In 2017, Uber's board commissioned former Attorney General Eric Holder and the law firm Covington and Burling to examine the company's culture and workplace practices, after a former employee publicly described sexual harassment at the company. The board unanimously adopted all 47 resulting recommendations in June 2017. They included adding an independent director, considering an independent chair, creating a larger and more independent audit committee, and reducing the chief executive's authority (Reuters, 2017). Travis Kalanick took a leave of absence and resigned as chief executive the following week (CNBC, 2017).

The Uber case shows how ethical and cultural failures become governance failures. The remedy was structural: independence on the board, a stronger audit committee, and limits on one executive's authority. Once the facts reached the board, its response was decisive. The harder question is why the facts reached it only after a former employee described them in public.

Board lesson. AI raises its own ethical questions: algorithmic bias, transparency, accountability, and effects on employment and privacy. Those questions are now being tested in court, and governance records are part of the evidence. In Estate of Lokken v. UnitedHealth Group, plaintiffs allege that the insurer used a predictive tool to override physicians' judgment and deny post-acute care; the company disputes that claim. In March 2026, a magistrate judge ordered discovery that reaches materials of the company's internal AI review board, including the identities of its members (Greenberg et al., 2026). I would not advise every board to form a separate ethics committee. I would advise every board to assign AI ethics to a committee by charter, to involve the people affected (clinicians, patients, members, and employees) in how AI is adopted, and to assume that its own records will one day be read by someone outside the boardroom.

AI: Expanding Business Strategies

The obstacles are significant, but AI also offers real opportunity to organizations that govern it well: efficiency, new products, competitive position, better insight, better service, and lower waste. The six opportunities below are the ones the original article named. Where the company examples it used could be validated, I have kept them as company facts; where they could not, I have used cases from the public record.

Enhancing Operational Effectiveness

AI can automate repetitive work and streamline processes. The strongest healthcare evidence comes from clinical documentation. A randomized trial of ambient AI documentation among 66 practitioners found that work exhaustion fell by 0.44 points on a five-point scale and note time fell by 0.36 hours per day, while documentation quality stayed high (Young, 2025). Revenue cycle work shows similar gains. Kaiser Permanente reports that AI-assisted appeal writing raised recovery rates by about 15 percent, and Nebraska Medicine cut staff time per appeal letter from about 15 minutes to between eight and ten (Williams, 2026).

Board lesson. The board should require management to define and report key performance indicators (KPIs) for each material AI initiative: efficiency gains, cost reductions, patient and staff experience, and, for any clinical use, safety and quality outcomes. Today most directors do not receive them. Only 18 percent of directors in the PwC survey rated the metrics linking AI outcomes, risks, and performance as good or excellent (Gaskin, 2026).

Developing Innovative Offerings: Apple, Microsoft, and Xerox

The original version of this article credited Xerox's Palo Alto Research Center (PARC) with the mouse. The record is more instructive. Douglas Engelbart introduced the mouse at the Stanford Research Institute (SRI) and filed for a patent in 1967. A former colleague, Bill English, brought the technology to PARC in 1971. Xerox first sold a mouse-equipped computer, the 8010 Star, in 1981. Apple made the mouse standard on the original Macintosh in 1984, and Microsoft Windows spread it widely in the 1990s (Scientific American, n.d.).

Like Kodak, the case shows that the organization that invents a technology does not always capture its value. The organizations that did were the ones that built it into products people would buy. Nvidia, which the original version of this article cited for gaming and autonomous vehicles, shows how far value can move. In its fiscal year ended January 25, 2026, data center products produced USD 193.7 billion of its USD 215.9 billion in revenue, against USD 16.0 billion from gaming and USD 2.3 billion from automotive (NVIDIA Corporation, 2026).

Board lesson. In life sciences, AI is moving from research tool to product. Insilico Medicine has launched a 320-patient Phase III trial of rentosertib for idiopathic pulmonary fibrosis, a drug the company describes as discovered and designed with its AI platform (Proffitt, 2026). The FDA has authorized more than 1,400 AI-enabled medical devices, including 331 in 2025, the most in a single year (Reuter & Han, 2026). Boards should support research and development in new AI applications, and should ask, for each one, who will capture the value and what evidence regulators and customers will require.

Maximizing Your Competitive Edge: Netflix

Netflix offered subscribers the option to watch films instantly on their computers in January 2007, while its disc rental business was still its core (Netflix, 2007). It chose to compete with its own business before a competitor did. It then invested in original content, releasing all 13 episodes of its original series House of Cards at once on February 1, 2013 (Baysinger, 2012). Salesforce made a comparable early move in its market, introducing Einstein in September 2016 to bring predictive lead scoring, deal insights, and recommendations into its customer relationship management (CRM) platform (Salesforce, 2016). The original article attributed this to the foresight of Netflix's board; I found no public source for that attribution, so the lesson here is about the company's choice, not its board.

Board lesson. Early adoption can create advantage, and in health care the competition is already under way between payers and providers. Hospitals and physicians faced an initial claim denial rate of 11.65 percent in 2025 through November, and providers are deploying AI to contest denials while payers deploy it to process claims (Williams, 2026). A board should ask where its organization stands in that contest, and whether the advantage it seeks depends on AI that its regulators, patients, or members would find acceptable.

Insights Driven by Data

AI and advanced analytics can reveal patterns in patient flow, demand, outcomes, and costs that people miss, and support better forecasts and decisions. The quality of those insights depends on the quality, security, and lawful use of the data behind them. Regulators have said so directly. In January 2026, the FDA and the European Medicines Agency issued ten joint principles of good AI practice in drug development, including a defined context of use, a risk-based approach, and data governance and documentation (Al-Faruque, 2026).

Board lesson. The board does not design data governance, but it should require one and ask how it is tested. Guidance from The Joint Commission and the Coalition for Health AI recommends data security standards, vendor data use agreements, and regular updates to senior leadership and the governing board on AI use, outcomes, and adverse events (Polsinelli, 2025).

Improving Customer Experience: Amazon

Amazon's recommendations are a well-documented example of personalization. Its engineers described an item-to-item method that matches each item a customer has bought or rated to similar items, which stays fast across a very large catalog (Linden et al., 2003). The original article credited Amazon's board with this capability; the public record credits the engineers.

Board lesson. AI can make the patient and member experience more personal, more efficient, and more convenient, through scheduling, navigation, and virtual assistants. In health care, personalization comes with duties of disclosure. Colorado's replacement law requires covered health care entities to give patients a general notice of their use of such technology (Darch et al., 2026). California's AB 1979 treats health care chatbot businesses as providers under the state's medical confidentiality law (Montano & Tran, 2026). The board should ask whether patients know when they are dealing with AI, and what happens when they object.

Promoting Sustainable Practices

AI can support sustainability by reducing waste. In the Lean Six Sigma discipline in which I trained, waste takes seven forms: inventory, motion, over-processing, overproduction, waiting, transport, and defects. Predictive maintenance can detect likely equipment failures before they occur and extend equipment life. The International Energy Agency (IEA) estimates that AI-led control of heating and cooling in buildings could save around 300 terawatt-hours (TWh) of electricity globally. The same report projects that data center electricity use will more than double, to about 945 TWh, by 2030 (International Energy Agency, 2025).

Board lesson. Sustainability matters to investors, regulators, patients, and employees, and AI cuts both ways. A board that reports AI's efficiency gains should also ask what the organization's AI use costs in energy and how its vendors report it.

AI: Expanding Role of the Board

Corporate boards have a growing responsibility to oversee their organizations through AI and the changes in technology, organization, and business it brings. That requires continuous learning, careful oversight, sound risk management, and a culture that welcomes innovation without hiding failure. The original article illustrated each area with a company board (Google, Amazon, Volkswagen, Johnson and Johnson, Apple, Procter and Gamble, and Patagonia). The public sources I could find support several of those company examples but not the claims about their boards. I have kept each example a source supports, stated what it shows, and supported the eight areas with current evidence.

Education and Awareness

Directors know they have a gap. In the PwC 2026 survey, 71 percent said their boards lack AI skills (Gaskin, 2026). Nonprofit health systems face the steepest climb. About 1,500 nonprofit and public hospital boards oversee about USD 900 billion in net patient revenue, yet only 32 percent require continuing education, and less than 0.01 percent of patient revenue goes to board development (Condon, 2026). Workshops, briefings, and expert advice help directors make informed decisions; the aim is fluency enough to ask the right questions, not technical expertise.

Guiding and Directing Strategy

Most boards expect AI to shape strategy. In a National Association of Corporate Directors (NACD) survey, 76 percent of directors said AI would factor into their 2026 growth strategy, while most organizations reported only modest gains from their AI investments to date (National Association of Corporate Directors, 2025). The board's role is to review AI strategy regularly, to test whether it serves the organization's mission and goals, and to promote collaboration between technology leaders and the rest of the business. Management writes the strategy; the board challenges and approves it. Amazon's first letter to shareholders, written when it was an online bookseller, committed the company to "make investment decisions in light of long-term market leadership considerations" rather than short-term profitability (Bezos, 1998). The original version of this article credited Amazon's board with that discipline. The record shows a commitment management made to shareholders, which is the kind of commitment a board can hold management to.

Managing Risks

AI risk is now a disclosure matter. The share of S&P 500 companies disclosing AI as a risk rose from 12 percent in 2023 to 83 percent in 2025, yet only 9 percent of surveyed executives said their companies were very prepared to comply with AI regulation (The Conference Board, 2026). Boards should oversee a risk management framework that covers AI, require regular evaluations and audits, and confirm that the framework changes as the technology does. In health care, The Joint Commission launched a voluntary certification for the responsible use of AI on June 1, 2026, which assesses governance, data management, risk and bias reduction, safety monitoring, and education (Guerra, 2026). Its standards give boards a ready reference for what to ask.

Guiding and Shaping the Talent and Culture

AI is already changing the workforce. Half of the healthcare organizations in the 2026 return-on-investment survey had cut headcount because of AI or planned to within six months, with reductions averaging 8 to 13 percent of the affected labor (Guerra & Kraus, 2026). Some companies now tie executive pay to AI goals; Recursion Pharmaceuticals weighted one at roughly 16.7 percent of its annual incentive (Batish & Gordon, 2026). Boards should ask for a workforce plan that covers recruitment, retention, upskilling, and redeployment, and should make sure any AI incentive rewards measured outcomes and safety rather than adoption volume. A culture of learning and reasonable risk-taking also requires that people can report AI failures without fear. Google offers an example of management treating talent as a question of evidence. Its people analytics group, formed in 2007, tested whether managers matter, found that even small improvements in manager quality were associated with lower turnover and higher satisfaction, and built the eight manager behaviors it identified into feedback surveys, training, and awards (Garvin, 2013). The original version credited Google's board; the source describes management's work.

Diversity of Mind

Boards have added little AI expertise. Disclosed AI expertise among S&P 500 directors rose only from 1.5 percent in 2021 to 2.7 percent in 2025, and 25 percent of surveyed executives rated their boards' AI fluency as low or none (The Conference Board, 2026). In my view, a board overseeing AI in health care benefits from directors with clinical, data, legal, and technology backgrounds, and from diversity of gender, ethnicity, and experience. That mix matters for AI in a specific way. I believe people with different experience are more likely to notice when a model treats patients or members differently. Johnson and Johnson shows what disclosure of that mix can look like. Its 2026 proxy statement reports, among 12 director nominees, healthcare industry experience for eight, science and technology for five, and digital experience for four, and it assigns emerging technologies, including AI, to the board's Science and Technology Committee (Johnson & Johnson, 2026). The original version credited the board's mix with the company's innovation; the proxy reports the composition, not its effect.

Working Together

AI oversight fails when it sits in one silo. The Joint Commission and Coalition for Health AI guidance recommends a multidisciplinary governance body, with clinical, technical, legal, and patient perspectives, integrated with existing compliance, risk, and patient safety structures (Polsinelli, 2025). Boards should promote that collaboration across departments, and should explore partnerships with technology providers, research institutions, and peer organizations, under the vendor standards described in the Nokia lesson. Apple offers a management example of collaboration by design. When Steve Jobs returned in 1997, he combined the separate functional departments of the business units into one functional organization, which Apple retains, and the authors of a study of that design credit it and Apple's leadership model with a central role in its innovation in hardware, software, and services (Podolny & Hansen, 2020). The original version credited Apple's board; the source describes how management organized the company.

Managing Knowledge

Continuous monitoring is how a board learns whether AI investments deliver what was promised. Directors should set metrics and benchmarks (efficiency gains, cost savings, patient and staff experience, safety, and innovation outcomes) and review them on a fixed schedule. They should also ask how the numbers are produced. The healthcare return figures cited in this article are self-reported multiples on money spent to date (Guerra & Kraus, 2026), and 66 percent of directors in the PwC survey rated the metrics linking AI to performance as fair or poor (Gaskin, 2026). Procter and Gamble shows the discipline at scale. It launched its Supply Chain 3.0 program, which combines AI with more basic automation, in 2023, tested it in pilots, and in 2026 moved it into full rollout against stated targets that include up to USD 1.5 billion in savings in cost of goods sold (Gonsalves, 2026). A board can ask management for the same sequence: a pilot, a stated target, and a report against it before the program scales.

Engaging With Stakeholders

Clinicians, patients, members, employees, suppliers, and shareholders all have a stake in how AI is used. Physicians want a voice: 85 percent want to be consulted on AI adoption, and 88 percent cite safety and efficacy validation as critical (American Medical Association, 2026). Investors want disclosure. Glass Lewis expects companies that use AI to disclose the board's oversight role and how directors stay informed, and it may recommend against directors where poor oversight has materially harmed shareholders (Gunderson Dettmer, 2024). Boards should establish regular channels for that feedback and show how it informs decisions. Some boards have written a stakeholder interest into governance itself. Since September 2022, a trust created to enshrine Patagonia's purpose and values has held all of the company's voting stock and elects and oversees its board, while a nonprofit that uses the company's profits to protect nature holds the nonvoting stock (Patagonia, 2022). The original version of this article described Patagonia's board working with environmental stakeholders; the public record shows a structure rather than a practice, but it shows how far a board can go in binding itself to the people its decisions affect.

The Strongest Objection: These Lessons Are Hindsight

A skeptical director could fairly say that these cases are hindsight. Kodak, Blockbuster, and Nokia made decisions that looked reasonable with the information available at the time. Netflix in 2000 was an unprofitable mail-order business. A touchscreen phone in 2004 was expensive and uncertain. If the lesson is "foresee the future," no board can meet it, and the cases teach nothing.

I agree with the premise, and that is why I draw a different lesson. The original version of this article asked boards to anticipate technology. That standard is impossible and, under Delaware law, it is not the standard. What the courts examine is whether directors made a good faith effort to put a reporting system in place for mission-critical risk, and whether they acted on what it told them (Marchand v. Barnhill, 2019). Wells Fargo and Teligent were not failures of foresight; they were failures of reporting. A board cannot guarantee that it will see the future. It can guarantee that it designed the path by which the present reaches it, and that its minutes show it used that path.

What the Six Cases Have in Common

None of these organizations lacked technology. Each lacked a dependable path for the right information to reach the board, or a board habit of asking for it. In my experience, the gap is rarely a lack of interest among directors. It is usually a reporting line that no one designed.

For AI, that path has four parts. The first is an inventory of where AI is used, including tools embedded in vendor products and tools staff adopted on their own. The second is a named accountable executive. The third is a committee charter that assigns oversight. The fourth is a set of conditions attached to every material AI approval: monitoring after launch, a date for reporting results, and criteria for pausing the system.

Technological change is a challenge for every board, and it is also an opportunity for organizations willing to invest and adapt. Boards that understand the obstacles and the possibilities, and that oversee their organization's approach to technology rather than run it, can help their organizations do well through this change. The priorities are the ones set out above: education, strategy, risk, talent, diversity, collaboration, monitoring, and engagement with the people AI affects.

For the Board

• Do we have a current inventory of every AI system in use, including those embedded in vendor products and those adopted without approval, and who is accountable for each?
• Which AI decisions require board or committee approval, and which conditions do we attach to those approvals?
• Do we receive reports on AI from compliance, risk, and internal audit, not only from the executives who sponsor the programs?
• How does management measure the return on each material AI investment, and are the figures self-reported?
• Do our vendor contracts address training on our data, notice of model changes, audit rights, and indemnification?
• Would our minutes show that we asked these questions, and what we decided?

Board question sets for each of these topics are in the Avancer AI Governance Navigator.

References

Al-Faruque, F. (2026, January 14). EMA, FDA issue joint AI guiding principles for drug developers. Regulatory Affairs Professionals Society. https://www.raps.org/resource/ema-fda-issue-joint-ai-guiding-principles-for-drug.html

American Medical Association. (2026, March 12). AMA: AI usage among doctors doubles as confidence in technology grows [Press release]. https://www.ama-assn.org/press-center/ama-press-releases/ama-ai-usage-among-doctors-doubles-confidence-technology-grows

Anthony, S. D. (2016, July 15). Kodak's downfall wasn't about technology. Harvard Business Review. https://hbr.org/2016/07/kodaks-downfall-wasnt-about-technology

Antioco, J. (2011). How I did it: Blockbuster's former CEO on sparring with an activist shareholder. Harvard Business Review, 89(4). https://hbr.org/2011/04/how-i-did-it-blockbusters-former-ceo-on-sparring-with-an-activist-shareholder

Associated Press. (2006). Kodak tops U.S. digital-camera market. NBC News. Retrieved October 21, 2026, from https://www.nbcnews.com/id/wbna11258928

Associated Press. (2021, March 26). VW seeks damages from former execs for diesel scandal. Transport Topics. https://www.ttnews.com/articles/vw-seeks-damages-former-execs-diesel-scandal

Batish, A., & Gordon, A. (2026, March 2). AI as a performance metric: What companies are disclosing now. Harvard Law School Forum on Corporate Governance. https://corpgov.law.harvard.edu/2026/03/02/ai-as-a-performance-metric-what-companies-are-disclosing-now/

Baysinger, T. (2012, October 4). Netflix to debut 'House of Cards' Feb. 1. Broadcasting & Cable. https://www.nexttv.com/news/netflix-debut-house-cards-feb-1-64651

Bezos, J. P. (1998, March 21). 1997 letter to shareholders. Amazon.com, Inc. https://www.aboutamazon.com/news/company-news/amazons-original-1997-letter-to-shareholders

Bloomberg. (2017, August 31). Wells Fargo increases fake account estimate 67% to 3.5 million. Fortune. https://fortune.com/2017/08/31/wells-fargo-increases-fake-account-estimate

Board of Governors of the Federal Reserve System. (2025, June 3). Federal Reserve announces Wells Fargo is no longer subject to the asset growth restriction from the Board's 2018 enforcement action against the bank [Press release]. https://www.federalreserve.gov/newsevents/pressreleases/enforcement20250603a.htm

Bonadurer, J. (2026, February). Delaware court denies dismissal of Caremark oversight claims in Teligent case, highlighting board and officer compliance obligations. Jenner & Block. https://www.jenner.com/en/news-insights/publications/delaware-court-denies-dismissal-of-caremark-oversight-claims-in-teligent-case-highlighting-board-and-officer-compliance-obligations

Brennen, S. B. (2026, July 6). Where state AI legislation stands half way into 2026. Tech Policy Press. https://techpolicy.press/where-state-ai-legislation-stands-half-way-into-2026

Chittum, R. (2014, January 7). Bank tellers in the boiler room. Columbia Journalism Review. https://www.cjr.org/?p=100376

CNBC. (2017, June 21). Eric Holder on Uber CEO Travis Kalanick's resignation. https://www.cnbc.com/2017/06/21/eric-holder-on-uber-ceo-travis-kalanicks-resignation-at-gateway-17.html

CNNMoney. (2012, January 19). Eastman Kodak files for bankruptcy. https://money.cnn.com/2012/01/19/news/companies/kodak_bankruptcy/

Condon, A. (2026, April 27). Health system governance is falling behind: 8 things to know. Becker's Hospital Review. https://www.beckershospitalreview.com/hospital-management-administration/health-system-governance-is-falling-behind-8-things-to-know/

The Conference Board. (2026, April 22). Report: In the S&P 500, disclosure of AI risks surges from 12% to 83% [Press release]. PR Newswire. https://digital-release.wowktv.com/business/press-releases/cision/20260422NY41234/report-in-the-sp-500-disclosure-of-ai-risks-surges-from-12-to-83

Consumer Financial Protection Bureau. (2016, September 8). Consumer Financial Protection Bureau fines Wells Fargo $100 million for widespread illegal practice of secretly opening unauthorized accounts [Press release]. https://www.consumerfinance.gov/about-us/newsroom/consumer-financial-protection-bureau-fines-wells-fargo-100-million-widespread-illegal-practice-secretly-opening-unauthorized-accounts/

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 & Gray. https://www.ropesgray.com/en/insights/alerts/2026/05/colorado-scales-back-ai-law-with-targeted-implications-for-health-care

Fox, A. (2024, September 20). Texas AG settles with clinical genAI company. Healthcare IT News. https://www.healthcareitnews.com/news/texas-ag-settles-clinical-genai-company

Garvin, D. A. (2013). How Google sold its engineers on management. Harvard Business Review, 91(12). https://hbr.org/2013/12/how-google-sold-its-engineers-on-management

Gaskin, J. L. (2026, September 23). Boards say they need more from management on AI. Corporate Compliance Insights. https://www.corporatecomplianceinsights.com/boards-say-they-need-more-from-management-ai/

Ghavi, A. R., Glucksman, H., & Katsuki, K. (2025, December 16). President Trump signs executive order challenging state AI laws. Paul Hastings. https://www.paulhastings.com/insights/client-alerts/president-trump-signs-executive-order-challenging-state-ai-laws

Gonsalves, A. (2026, May 8). P&G shifts Supply Chain 3.0, other platforms into large-scale rollout. Supply Chain Dive. https://www.supplychaindive.com/news/pg-shifts-supply-chain-30-other-platforms-into-large-scale-rollout/819256/

Goodrich, J. (2022, April 6). The first digital camera was the size of a toaster. IEEE Spectrum. https://spectrum.ieee.org/first-digital-camera-history

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

Gunderson Dettmer. (2024, November 21). Glass Lewis releases 2025 benchmark policy updates. https://www.gunder.com/en/news-insights/blogs/public-ventures/glass-lewis-releases-2025-benchmark-policy-updates

Hardawar, D. (2010, September 27). Nokia had iPhone-like prototype in 2004, but killed it. VentureBeat. https://venturebeat.com/business/nokia-had-iphone-like-prototype-in-2004-but-killed-it

Hughes, J. (2026, September 17). Healthcare's agentic AI boom is outpacing security, governance: Report. Healthcare Dive. https://www.healthcaredive.com/news/healthcares-agentic-ai-boom-is-outpacing-security-governance-report/830697/

Hunton Andrews Kurth. (2026, July 28). EU Digital Omnibus on AI enters into force. Privacy and Cybersecurity Law Blog. https://www.hunton.com/privacy-and-cybersecurity-law-blog/eu-digital-omnibus-on-ai-enters-into-force

In re Caremark International Inc. Derivative Litigation, 698 A.2d 959 (Del. Ch. 1996).

International Energy Agency. (2025). Energy and AI: Executive summary. Retrieved October 21, 2026, from https://www.iea.org/reports/energy-and-ai/executive-summary

Janisch, A., Gerhardt, W., & Shukla, M. (2025, December 11). 2026 US health care outlook. Deloitte Insights. https://www.deloitte.com/global/en/insights/industry/health-care/life-sciences-and-health-care-industry-outlooks/2026-us-health-care-executive-outlook.html

Johnson & Johnson. (2026, March 11). Notice of 2026 annual meeting and proxy statement (Schedule 14A). U.S. Securities and Exchange Commission. https://www.sec.gov/Archives/edgar/data/200406/000020040626000063/jnj-20260309.htm

Kohn, A. H., & Marcogliese, P. L. (2017, May 16). With the benefit of hindsight: The Wells Fargo sales practices investigation report. Harvard Law School Forum on Corporate Governance. https://corpgov.law.harvard.edu/2017/05/16/with-the-benefit-of-hindsight-the-wells-fargo-sales-practices-investigation-report

Lev-Ram, M. (2011, February 11). NokSoft born: Nokia moves to Windows, takes on Google and Apple. Fortune. https://fortune.com/2011/02/11/noksoft-born-nokia-moves-to-windows-takes-on-google-and-apple

Linden, G., Smith, B., & York, J. (2003). Amazon.com recommendations: Item-to-item collaborative filtering. IEEE Internet Computing, 7(1), 76–80. https://www.cs.umd.edu/%7Esamir/498/Amazon-Recommendations.pdf

Marchand v. Barnhill, 212 A.3d 805 (Del. 2019). https://law.justia.com/cases/delaware/supreme-court/2019/533-2018.html

Mollman, S. (2023, April 14). Netflix cofounder Marc Randolph recalls Blockbuster rejecting chance to buy it. Fortune. https://fortune.com/2023/04/14/netflix-cofounder-marc-randolph-recalls-blockbuster-rejecting-chance-to-buy-it

Montano, M., & Tran, S. (2026, October 8). Newsom approves 3 health care AI bills, vetoes 2. California Health Care Foundation. https://chcf.org/blog/2026/10/08/newsom-approves-3-health-care-ai-bills-vetoes-2

National Association of Corporate Directors. (2025, December 11). Boards prioritize strategic execution, technology and people heading into 2026 [Press release]. https://www.nacdonline.org/about/newsroom/press-release/press-release/boards-prioritize-strategic-execution-technology-people-2026/

Netflix. (2007, January 16). Netflix offers subscribers the option of instantly watching movies on their PCs [Press release]. https://about.netflix.com/en/news/netflix-offers-subscribers-the-option-of-instantly-watching-movies-on-their

NVIDIA Corporation. (2026). CFO commentary on fourth quarter and fiscal 2026 results (Form 8-K exhibit). U.S. Securities and Exchange Commission. https://www.sec.gov/Archives/edgar/data/1045810/000104581026000019/q4fy26cfocommentary.htm

Patagonia. (2022, September 14). Patagonia's next chapter: Earth is now our only shareholder [Press release]. https://www.patagoniaworks.com/press/2022/9/14/patagonias-next-chapter-earth-is-now-our-only-shareholder

Podolny, J. M., & Hansen, M. T. (2020). How Apple is organized for innovation. Harvard Business Review, 98(6). https://hbr.org/2020/11/how-apple-is-organized-for-innovation

Polsinelli. (2025, October 29). Joint Commission and Coalition for Health AI release first-of-its-kind guidance on responsible AI use in healthcare. https://www.polsinelli.com/publications/tjc-chai-release-guidance-responsible-ai-use-healthcare

Pressly, D. F., & Leeds, E. I. (2025, April 17). Section 1557: Deadlines approach, but long-term prognosis unclear under executive orders. Ballard Spahr. https://www.ballardspahr.com/insights/alerts-and-articles/2025/04/section-1557-deadlines-approach-but-long-term-prognosis-unclear-under-executive-orders

Proffitt, A. (2026, July 8). Insilico Medicine launches Phase III trial for AI-developed drug. Clinical Research News. https://www.clinicalresearchnewsonline.com/news/2026/07/08/insilico-medicine-launches-phase-iii-trial-for-ai-developed-drug

Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data (General Data Protection Regulation). (2016). Official Journal of the European Union, L 119. https://eur-lex.europa.eu/eli/reg/2016/679/oj

Reuter, E., & Han, J. Y. (2026, May 27). AI in medtech is booming. Track new devices here. MedTech Dive. https://www.medtechdive.com/news/ai-medtech-track-new-devices-fda/748397/

Reuters. (2017, June 14). Uber CEO takes leave of absence amid sweeping changes after scandals. Business Insurance. https://www.businessinsurance.com/uber-ceo-travis-kalanick-leave-of-absence-sweeping-changes-after-scandals/

Reuters. (2020, March 17). Volkswagen says diesel scandal has cost it $34.69B. Business Insurance. https://www.businessinsurance.com/volkswagen-says-diesel-emissions-scandal-has-cost-it-3469b-frank-witter/

Ricker, T. (2010, September 23). Blockbuster files for Chapter 11 bankruptcy. Engadget. https://www.engadget.com/2010-09-23-blockbuster-files-for-chapter-11.html

Salesforce. (2016, September 19). Salesforce introduces Salesforce Einstein: Artificial intelligence for everyone [Press release]. https://investor.salesforce.com/news/news-details/2016/Salesforce-Introduces-Salesforce-Einstein---Artificial-Intelligence-for-Everyone/default.aspx

Schroeder, P. (2023, March 15). Former Wells Fargo executive pleads guilty to obstructing bank examination, fined $17 mln in fake accounts scandal [Reuters]. Kitco News. https://www.kitco.com/news/off-the-wire/2023-03-15/former-wells-fargo-executive-pleads-guilty-obstructing-bank

Scientific American. (n.d.). The origin of the computer mouse. Retrieved October 21, 2026, from https://www.scientificamerican.com/article/origins-computer-mouse/

State of California Department of Justice. (n.d.). California Consumer Privacy Act (CCPA). Office of the Attorney General. Retrieved October 21, 2026, from https://oag.ca.gov/privacy/ccpa

U.S. Environmental Protection Agency. (2017, January 11). Volkswagen agrees to plead guilty, pay $4.3 billion in criminal and civil penalties [Press release]. https://www.epa.gov/archive/epa/newsreleases/volkswagen-agrees-plead-guilty-pay-43-billion-criminal-and-civil-penalties.html

Washington Times. (2017, January 11). VW pleads guilty in emissions scandal. https://www.washingtontimes.com/news/2017/jan/11/vw-pleads-guilty-emissions-scandal/

Wells Fargo & Company. (2017, April 10). Wells Fargo board releases findings of independent investigation of retail banking sales practices and related matters [Press release]. https://newsroom.wf.com/press-release/community-banking-and-small-business/wells-fargo-board-releases-findings-independent

Williams, J. (2026, January 30). Battle of the bots intensifies over denials. Healthcare Financial Management Association. https://www.hfma.org/revenue-cycle/denials-management/battle-of-the-bots-intensifies-over-denials/

Young, K. (2025). Ambient AI scribes linked to lower work exhaustion in multistate pragmatic trial. Consultant360. Retrieved October 21, 2026, from https://www.consultant360.com/exclusive/ambient-ai-scribes-linked-lower-work-exhaustion-multistate-pragmatic-trial