Weekly Intelligence for Directors Who Want to See What's Coming

February 18, 2026 | Issue #3

Fifty-six percent. That's the share of organizations reporting neither increased revenue nor reduced costs from their AI implementations over the past twelve months, according to PwC's 29th Global CEO Survey, released in January at the World Economic Forum. The survey polled 4,454 chief executives across 95 countries. More than half came away empty-handed.

The numbers get worse from there. An MIT study of enterprise generative AI investment found that 95% of organizations are achieving zero measurable return on an estimated $30 to $40 billion in cumulative spending. The IBM Institute for Business Value reported that enterprise-wide AI initiatives achieved an ROI of just 5.9%, despite requiring a 10% capital investment. Gartner expects spending on AI application software to nearly triple to roughly $270 billion in 2026. And Kyndryl's Readiness Report, drawing on insights from 3,700 business executives, found that 61% of CEOs say they are under increasing pressure to show returns compared with a year ago.

This is no longer an innovation story. It is becoming a governance story.

The fiduciary question boards need to confront

There is a meaningful difference between "patient capital" and "unexamined spending." Boards have always understood that transformational investments take time to mature. But the AI investment cycle has a feature that prior technology cycles did not: the costs are accelerating while the returns remain theoretical for most organizations.

Per-employee AI tool costs now range from $590 to $1,400 annually, according to internal data shared with Fortune by Lexi Reese, a former strategy executive who has had conversations with over 300 customers about their AI spending. For a company with 10,000 employees, that is $6 million to $14 million per year in tool costs alone, before infrastructure, talent, and integration expenses. Multiply that across a multi-year commitment and the capital at risk becomes material by any board's standards.

The question is not whether AI will eventually deliver value. Most governance experts believe it will. The question is whether boards are applying the same rigor to AI spending that they would apply to any other capital allocation decision of comparable scale. For many, the answer is no.

Forrester's Brian Hopkins, vice president of emerging technology, told CIO magazine that about half of the organizations in his firm's client base are putting off planned AI outlays in 2026 because they cannot connect task-level efficiency gains (saving 15 minutes on an email) to process-level improvements that show up in earnings. "Because firms are having a hard time moving from individual task efficiency into more process efficiency, they're not seeing any earnings at the top line," Hopkins said.

Where the liability line is moving

Three developments are shifting AI spending from a management prerogative to a board-level fiduciary exposure.

First, regulators are beginning to frame the absence of quantifiable AI metrics as a failure of oversight. According to a January 2026 analysis by OnBoard, failure to implement quantifiable metrics for AI performance and risk is being treated by regulators as a failure of the board's duty of oversight, particularly in the finance and healthcare sectors. This echoes the Caremark standard, which establishes that directors breach their duty of care when they fail to make a good faith effort to oversee a corporation's operations.

Second, proxy advisors and institutional investors are updating their expectations. A November 2025 Corporate Board Member briefing reported that BlackRock, Allianz, and Glass Lewis have already updated stewardship guidelines, with ISS close behind. By the 2026 proxy season, they expect boards to demonstrate AI literacy and document director training and oversight frameworks in proxy statements. Fall short, and directors could face withhold recommendations, reputational damage, or Caremark-style derivative claims.

Third, the Equal AI/WilmerHale Governance Playbook, published in January, stated directly: "AI governance isn't just good practice; it has quickly become a legal and strategic imperative. Boards that take the time now to assess governance structures and elevate AI literacy will meet Delaware's oversight standards." The implication of the inverse is hard to miss.

What separates boards that are governing AI from those that are approving budgets

The distinction is straightforward. A board that is governing AI spending asks five questions that a board merely approving budgets does not:

The fifth question may be the most important. McKinsey found that nearly 70% of Fortune 500 companies are using Microsoft 365 Copilot, but 92% of executives plan to increase AI spending in the next three years. Meanwhile, only 15% of US employees report that their workplaces have communicated a clear AI strategy, according to Gallup. When spending is rising while strategy remains unclear to the people executing it, the board should be concerned.

The path from here

None of this argues against AI investment. It argues for AI governance. The companies that are seeing returns, which PwC, MIT, and McKinsey all identify as a small but growing minority, share common traits: executive-level strategy tied to specific business outcomes, centralized deployment through what PwC calls an "AI studio," disciplined use-case selection, and rigorous measurement from the start.

For boards, the action is not to slow AI spending. It is to ensure that AI spending is subject to the same oversight discipline as every other material capital allocation. That means defined success metrics, kill criteria for underperforming initiatives, clear reporting cadences, and honest assessment of whether the organization has the data infrastructure, talent, and change management capacity to make its AI investments productive.

The 56% of companies seeing no returns are not necessarily making bad bets. Many are making ungoverned ones. And that distinction is where board liability begins.

Questions to ask in your next board meeting:

→ "What is the total annualized cost of our AI program across all categories: tools, infrastructure, talent, integration, and opportunity cost? How does that compare to what was budgeted?"

→ "Of our active AI initiatives, how many have defined success metrics with measurable baselines? How many have met their targets?"

→ "What is our framework for terminating AI projects that do not demonstrate returns within an agreed timeframe? Has that framework ever been triggered?"

→ "Are we investing in AI because we have identified specific, high-value use cases tied to business outcomes, or because we are responding to competitive pressure and vendor enthusiasm?"

Five signals board members should track this week

1. ISS Shifts to Case-by-Case on ESG Shareholder Proposals

ISS Governance's 2026 Benchmark Policy Updates, effective for stockholder meetings on or after February 1, replace a previous presumption of support for many common environmental and social proposal topics with a case-by-case approach. This applies to climate risk reporting, greenhouse gas reporting, diversity practices, human rights reporting, and political contributions disclosure. The change is widely seen as responsive to political pressure and declining support levels at annual meetings. For directors: your proxy season playbook from last year may no longer apply. Boards should review their engagement strategies and anticipate less predictable voting outcomes on E&S proposals this cycle.

2. The Edelman Trust Barometer Signals a Governance Risk

The 2026 edition of the Edelman Trust Barometer, one of the most widely cited global measurements of institutional trust, has landed with a message that governance professionals should take seriously. At a time when boards are navigating AI, regulatory fragmentation, and geopolitical volatility, the state of public trust in institutions is a variable that affects license to operate, talent retention, and shareholder engagement. Directors should review the findings in the context of their company's stakeholder relationships.

3. Over 75% of Companies Did Not Update Risk Factor Disclosures After Tariff Announcements

A review of the first quarterly reports filed after April 2, 2025, found that over 75% of companies did not update their risk factor disclosures to reflect material changes since their annual reports, despite significant shifts in trade policy, government funding, and corporate sustainability reporting. Only 94 companies filed updated risk factors. For directors: if your company's risk factors do not reflect the tariff and trade policy environment your business is actually operating in, you have a disclosure gap that plaintiffs' counsel will notice.

4. Delaware Court Reverses Earnout Decision in Johnson & Johnson v. Fortis Advisors

On January 12, 2026, the Delaware Supreme Court issued an en banc opinion reversing the Court of Chancery's decision that Johnson & Johnson breached its efforts obligations regarding an earnout payment in its 2019 acquisition of medical robotics company Auris Health. The ruling refines how courts evaluate "efforts" clauses in merger agreements, a perennial source of post-closing disputes. For directors: if your company is party to an acquisition agreement with earnout provisions, this decision may affect how your obligations are interpreted. Work with counsel to review.

5. Texas Business Court Attracting Filings From Delaware Incorporations

As of January 2026, the newly established Texas Business Court is seeing an influx of filings from companies relocating from Delaware. This follows legislative changes intended to make Texas a more board-friendly jurisdiction. Companies are increasingly conducting "jurisdictional audits" to determine whether moving their state of incorporation would provide better protection against shareholder derivative suits. Combined with Delaware's SB 21 reforms (covered in Issue #2), the competitive landscape for corporate governance law is shifting. Directors should understand how these jurisdictional dynamics affect their own liability exposure.

Each week, one question worth raising at your next meeting.

"If our largest AI initiative failed to deliver measurable ROI within 18 months, would this board know, and would we have a predetermined framework for what to do about it?"

This question tests two governance capabilities simultaneously. First, whether the board has visibility into AI performance against defined benchmarks, not just management's narrative about "progress" and "momentum." Second, whether there is a decision framework in place before it is needed. Boards that establish kill criteria and escalation triggers before an initiative underperforms are governing. Boards that discover performance problems only when they become material enough for management to disclose them are not.

What's moving in Washington and beyond

Commerce Department AI Review (Update)

The administration's evaluation of state AI laws, first noted in Issue #2, remains pending. The review could recommend federal preemption of state laws like Colorado's AI Act, or leave the current patchwork intact. Either outcome has significant compliance implications. No updated timeline has been published.

California SB 253 Climate Reporting: August 10 Deadline

First reporting requirements under the Climate Corporate Data Accountability Act take effect August 10. Companies doing business in California, regardless of headquarters location, must oversee the collection of Scope 1 and 2 greenhouse gas emissions data. As noted in Issue #2, legal challenges remain unresolved, but compliance planning should not wait for judicial resolution.

ISS 2026 Policy Changes Now in Effect

ISS Benchmark Policy Updates took effect February 1 for the current proxy season. Beyond the ESG proposal changes noted above, updates include new approaches to executive compensation responsiveness (allowing more flexibility for companies that demonstrate good-faith shareholder engagement) and a new policy on dual-class share structures that will trigger "Withhold" or "Against" recommendations for directors unless reasonable sunset provisions are included.

A note on sequence: we had planned to examine government equity stakes as a governance variable this week. As the AI ROI reckoning has accelerated, with new data from PwC, MIT, and Forrester all landing in recent weeks, we prioritized this topic. The government-as-shareholder analysis remains on our list and we will revisit it when developments warrant.

Next week: Your board has an AI committee problem. Most boards still route AI oversight through audit. But with AI touching strategy, workforce, risk, compliance, and capital allocation simultaneously, the committee structure at most companies is not built for the task. We will look at what governance experts are recommending, what early adopters are doing, and whether your board's charter needs rewriting.

Researched, written, and edited in collaboration with Claude by Anthropic.