In October 2026, big tech stocks lost approximately $2 trillion in market value. Investors reacted out of fear that the AI bubble was over. The selloff showed how quickly confidence can shift when a compelling technology story meets an uncomfortable question about cost and missed profits.
That does not prove AI is a bubble, nor does it establish that all big tech companies will fail. A falling share price is not, by itself, evidence of poor oversight. It does, however, offer every board a reason to examine the assumptions supporting its own AI strategy.
Directors cannot prevent markets from repricing uncertainty. They can demand that management distinguish a promising technology from a defensible business model. The essential boardroom response is governance: disciplined oversight of where capital goes, which risks the company accepts, and what evidence would justify changing direction.
Challenge the AI Economics Before Approving the Spending
The back-and-forth in AI layoffs at Amazon and Google, along with the recent market correction, mattered because it questioned an investment thesis: that increasingly capable AI requires enormous spending, while simultaneously enabling employees to do more with less. Big tech companies are building an enormous number of data centers. At the same time, they are repeatedly destroying their company culture by cutting their employee base. Yet they have nothing to show for it. Microsoft makes still more money from traditional software. Google’s AI spending has reduced its AD revenue. So the economics don’t pan out.
Boards must treat these results seriously. Before approving significant AI expenditure or fundamentally altering an organization, directors should ask management to explain the economics under less favorable conditions. What happens if customers will not pay the expected premium? What happens if comparable capabilities become cheaper? Can the company reduce spending without undermining its core operations?
The objective is not to forecast every disruption. It is to avoid approving an investment that makes sense only while its most optimistic assumptions remain untouched.
Stop Confusing AI Adoption With Business Value
An AI deployment is not a business outcome. Neither is a successful demonstration. Boards should insist on knowing which operational problem an initiative addresses and how management will determine whether the investment worked. Meaningful AI use cases connect to measurable business objectives, with attention to both benefits and consequences when systems fail.
For directors, that means moving beyond reports about licenses purchased or employees trained. A customer-service deployment should demonstrate improvements in resolution quality and total cost, not merely more automated interactions. A coding assistant should be evaluated against delivery quality and review effort, not just generated output.
Management should establish those measures before scaling. Otherwise, the board risks approving additional spending because the organization has become attached to the initiative, not because the evidence supports expansion.
Accountability Goes Where Decisions Happen
At the same time, AI oversight extends beyond boards and the technology department. Procurement selects vendors, business leaders redesign workflows, legal teams assess exposure, and finance approves expenditure. Without explicit ownership, each function can assume another has considered the consequences.
The National Institute of Standards and Technology’s AI Risk Management Framework calls for documented responsibilities and clear communication around identifying, measuring, and managing AI risks. Its governance guidance also describes integrating AI oversight into broader enterprise risk management.
Boards should translate that principle into an accountable operating structure. Material initiatives need an executive owner with authority to address cross-functional problems. Directors need reporting that connects performance, expenditure, incidents, and unresolved risks.
The board must also settle its responsibilities. Whether oversight sits with the full board or a committee matters less than having a clear mandate. An impressive committee title cannot compensate for information that arrives too late to influence a decision.
Govern the AI Claims as Carefully as the Technology
The pressure to appear competitive creates another governance problem: public statements can move faster than operational reality.
Already in March 2024, the U.S. Securities and Exchange Commission announced settlements with investment advisers Delphia and Global Predictions over false and misleading statements about their purported AI use. The firms agreed to pay combined civil penalties of $400,000 without admitting or denying the findings.
Those cases involved investment advisers, not a general ruling on every corporate AI statement. The broader governance lesson is nevertheless straightforward: claims need evidence.
Boards should expect appropriate review of material statements about AI capabilities and financial benefits, especially when releasing them in their K-1s or quarterly statements. Management should distinguish an experiment from a production system, and estimated savings from realized results. Legal review should connect with technical verification, rather than operate on a separate track.
A confident claim may attract attention today. If the company cannot substantiate it tomorrow, that attention becomes a liability.
Build the Ability to Change Direction
Better governance should preserve room to innovate, not require certainty before experimentation. The practical approach is to make commitments proportionate to the evidence.
Boards can set the framework for bounded pilots, and bind expenditure plans to demonstrated results. They can also require management to explain how to pause or replace an initiative and how it manages the associated risks. In risk management, vendor dependence deserves particular attention. Recent market corrections may mean tools become unavailable at a moment’s notice if the companies behind them run out of money. Prudent governance guidance emphasizes sufficient control over third-party tools, contractual safeguards, and ongoing performance monitoring.
Directors should also make disagreement useful. The overall company culture should take warnings from finance, security, or frontline employees seriously. These warnings should be examined carefully, and not dismissed as resistance to innovation.
The next AI market shock may arise from cheaper models, disappointing returns, or something nobody anticipated. Boards do not need to predict its cause. They need a company that can respond without defending yesterday’s assumptions at any cost. That is the standard against which AI oversight should be judged.
Leave a Reply