Board members, start asking the right questions

AI is no longer only a technology adoption issue. It is an operating model issue, a workforce issue, a margin issue, a customer expectation issue, and a competitive strategy issue all at once. It is changing how organizations produce work internally. It is changing what customers expect externally. A company may become more efficient and still lose, because competitors used the same window to reimagine the product, the service, the advice or the experience itself.

That is the part boards should sit with.

AI will not simply make old processes faster. Often, it will reveal that the process was the problem. KPMG Canada's research found that 93 per cent of Canadian business leaders say their organizations use generative AI in some form, but only 31 per cent have integrated it across operations and workflows, and only two per cent report measurable ROI. The issue is not lack of activity. It is the gap between experimentation and value. Deploying licences is using AI. Re-engineering an entire process to leverage the speed and accuracy of intelligence is using AI effectively. The two per cent are doing the latter. Boards should focus on strategic redesign, not traditional metrics like AI penetration or user counts.

We encourage boards to ask, debate and answer the following questions:

• What would this workflow look like if it were designed today with AI?
• What new products, services or client experiences become possible when AI is embedded directly into the way the organization operates?
• Which decisions could be faster, better informed or more consistent?

KPMG and INSEAD collaborated on a framework for boards to help answer these questions. Through consultation with board members and specialists globally, it became clear that the dilemmas and trade-offs fall naturally into three perspectives: company, ecosystem and internal board practices.

Boards should not manage AI or replace executive judgment. They must engage deliberately with the strategic and transformational dilemmas AI creates, while remaining within their oversight role.

These three perspectives are addressed through the following five principles:

First, strategic oversight for long-term value. AI's upside is real, but realizing it often requires early investment in technology, data infrastructure, capability building and change management. Boards must balance short-term gains with longer-term innovation, assessing AI costs, benefits and risks against values and risk appetite. The core question is not which AI initiatives to approve, but how AI changes the firm's theory of value. When the artifact, analysis or advice a company once sold becomes cheaper, faster and more abundant, what is the company worth a premium for? Efficiency is the floor, not the ceiling. AI is reshaping service delivery, product development, customer expectations and business model defensibility.

Second, active technology and security oversight. Directors need not become machine-learning engineers, but they must understand management's architectural choices: what is built or bought, what data is used, which vendors are embedded, and where the firm depends on systems it does not control. If a model shaping pricing, underwriting, hiring or clinical decisions is owned and tuned elsewhere, the company rents part of its judgment. Technology sovereignty is now a governance issue and cybersecurity, privacy and AI security risks are enterprise-level.

Third, workforce transformation and human accountability. AI will change roles, skills, team structures and the value of expertise. Boards should not ask only how many hours AI may save. They should ask what capabilities the organization needs next, and what workforce can build them. They must also draw a clear line: AI can inform, but it cannot own. A model may generate a recommendation, but the firm, its officers, employees and directors own the consequence.

Fourth, building trustworthy AI. Trust is not a communications exercise after deployment. It must be engineered into governance from the outset through data quality, documentation, model oversight, red-teaming, testing, monitoring, human review, cybersecurity controls, privacy protections and escalation paths. Boards should reject the false choice between speed and governance. Governance is not the brake; it is what makes speed trustworthy. The board oversees whether management is deploying AI with fairness, transparency, safety, inclusivity and accountability, reflecting values and strategy.

Fifth, the work of the board itself. AI does not only change the company. It changes what boards need to know, and how often they need to know it. Should the board recruit deeper technical expertise? Create a technology or AI committee? Should the CEO present the AI strategy? What metrics should directors expect each quarter?  Boards need new cadences, sharper briefings and more dynamic decision-making to govern at the speed AI requires.

Some boards will be tempted to solve all of this by adding a single AI expert. AI experts can certainly help. But expertise without governance architecture will not be enough. Others will create a technology committee. That may also help. But if the AI conversation becomes siloed from strategy, workforce, risk and capital allocation, the board has simply moved the problem into a smaller room.

The better approach is to make AI visible as a whole-board issue. Management should provide a current view of where AI is used, what value it should create, what risks it introduces, what decisions it touches, and what remains human-owned. Boards should expect a corporate AI report: not compliance, but a living account of how AI is changing the business.

The companies that get the most from AI will not have the most pilots. They will challenge how work should happen.

That is where the opportunity becomes exciting. Not low-hanging fruit. Not productivity theatre. Not tools layered onto yesterday's processes. The opportunity is asymmetric value: hard problems redesigned end-to-end, faster insight generation, smarter risk and compliance, better knowledge systems, new customer experiences, and operating models that make organizations more capable.

The next phase of AI will be defined by who can redesign work, protect trust, preserve human accountability and create value while the ground is still moving.

That is the work of the board now.