Boards Lack Oversight of AI Talent and Skills, WTW Survey Finds

Only one in four directors believes their board understands the workforce effects of artificial intelligence well enough to oversee them, even though four in five call human capital critical to AI advantage, a WTW survey published on November 4 found.

The result puts board oversight of AI under a different kind of scrutiny. Most AI governance work in banks and large companies centers on model risk, data quality and cybersecurity. The survey points to a gap in the people side: who has the skills, whether the culture supports adoption and whether the board gets data it can trust.

WTW ran the survey with Directors & Boards and built on a similar study from 2024. The 2026 edition ranks human capital management among boards’ top three priorities, alongside purpose and strategy development and financial performance monitoring and reporting.

Board oversight of AI: where directors see gaps

Boards’ current AI priorities center on operational efficiency and cost reduction, followed by productivity and decision making. Over the next three years, directors expect AI to move beyond replacing tasks toward new sources of value, such as better customer experience and stronger commercial capabilities.

Almost two-thirds of directors cite data quality, privacy and cybersecurity vulnerabilities as the most significant challenges to enterprise AI adoption. About half name cultural change, workforce impact and a lack of critical AI skills as key risks.

Yet cultural change, building organizational AI skills and workforce impact rank near the bottom of boards’ current AI priorities. Directors identify the risk but do not put it on the agenda.

The agenda figures are sharper. Some 64% of directors do not believe boards spend enough time discussing investment in AI and critical skills. A further 57% say their boards do not receive enough information from management to oversee that investment effectively.

The authors, Kenneth Kuk, John Bremen and Saavaiz Sohail of WTW, put it this way: “Directors are not short on appetite, but they are short on airtime and data they can trust.”

Directors ranked purpose and strategy first among areas where they want more boardroom time, followed by human capital and AI. Those ranked ahead of capital investments and M&A, enterprise risk management and financial performance.

Why weak data limits board oversight of AI

For compliance, risk and governance teams, the information problem is the one they can most directly affect. They often supply the dashboards and board reports that directors rely on.

WTW’s authors link the data shortfall to underdeveloped measurement. They point to studies finding that most companies do not see AI’s impact on enterprise earnings or growth, and that most CEOs have not yet seen significant financial benefits from AI.

They describe an AI value capture gap. Measurement and incentive systems often track activity, such as usage rates and tokens consumed, rather than outcomes such as margin expansion from productivity gains or growth from stronger commercial capabilities.

A board that sees only usage statistics cannot judge whether an AI program works or whether it is creating risk. That bears on board compliance reporting, where the question is which indicators reach directors and how often.

Four challenges directors raised in the roundtable

WTW and Directors & Boards convened members of the Directors Thinking Ahead Group in September 2026 to test the survey results. The discussion centered on human factors rather than technology. Directors named four practical challenges:

• Culture can slow AI adoption when employees do not understand their role in the company’s AI vision or fear for their jobs.
• Productivity measures can overstate gains if faster workflows reduce quality, and middle managers may spend more time reviewing outputs and escalating concerns.
• Successful AI adoption depends on human judgment, yet existing management and leadership development programs may not teach people how to apply it.
• Restricting access to AI tools to limit privacy and confidentiality risk may backfire if employees use public platforms through personal accounts, which raises the risk of company information entering public models.

The last point matters for control owners. A ban on approved tools can push use out of sight, which is harder to monitor than a managed deployment.

Full board or committee: who owns AI risk

Consistent with the survey, most roundtable directors believe oversight of AI and human capital strategy belongs to the full board, with specific issues delegated to the appropriate committees. AI touches risk, capital allocation, technology, regulatory compliance, leadership and workforce capabilities, so governance requires a holistic view.

Directors noted that assigning AI oversight to a technology committee alone may fragment boardroom discussion and slow decisions while AI changes quickly. That is a live design question for firms that have given AI to a single risk or technology committee. The wider governance debate is covered in AI Without Oversight: The New Fault Line in Corporate Governance.

Directors also asked whether boards should have AI expert directors. They said the question will grow as regulatory frameworks take shape and institutional investors ask more about AI governance. The roundtable did not reach an answer.

Detectability, resilience and the harness

The roundtable also added a third dimension to risk assessment. Traditional frameworks weigh probability and severity. Drawing on systems engineering, directors proposed adding detectability: how hard it would be to identify a risk before it causes harm.

Members said boards must look beyond static, backward-looking dashboards. Because not every AI risk can be predicted or prevented, they viewed organizational resilience as an imperative. Boards should expect management to recognize disruption, respond and recover quickly, alongside conventional controls rather than in place of them.

Directors further observed that boards and management focus too much on the AI models themselves and overlook the harness. WTW defines this as the orchestration layer that gives models the organization-specific context and data to perform tasks. Permissions, compliance and auditability reside there, which makes it central to oversight.

That is a point internal audit and control teams can act on. Auditing the model alone leaves out the layer where access rights and logging are set.

The regulatory link is narrower than the survey

The survey does not test legal duties, and its findings should not be read as a compliance finding. WTW’s authors say the board’s duty of care over AI and human capital can no longer be exercised separately, but they do not tie that to a specific statute.

The nearest hard rule on skills is Article 4 of the EU AI Act, Regulation (EU) 2024/1689. It requires providers and deployers of AI systems to take measures to ensure a sufficient level of AI literacy among their staff and others operating AI systems on their behalf. The text is on EUR-Lex, and the literacy duty has applied since February 2, 2025.

That duty sits at the organizational level and does not name directors. Firms that also use systems classed as high risk face further obligations, set out in the EU AI Act high-risk pack. How far the board itself needs AI skills is, for now, a matter of governance practice and investor expectation.

WTW’s June 2026 analysis of risks shaping director and officer agendas also lists AI among the risks rising on boards’ agendas.

Questions the authors put to boards

The article closes with six questions boards can use. They ask how AI is being used across the business, whether the board measures AI value through outcomes rather than activity, and whether oversight covers both models and the harness.

They also ask how the board keeps pace with AI development, how management is preparing for adverse AI events, and how the board would know if something was going wrong. The last includes what is monitored, who is alerted and whether the board has tested management’s assurances by asking how they know.

The survey’s sample size, respondent mix and regional split are not set out in the published article, so the percentages describe the directors WTW polled and not boards generally. The 2024 study provides the comparison point for any follow-up. The questions directors raised on AI experts and investor pressure are likely to return as proxy season approaches and the EU AI Act’s further obligations phase in.