Boards are asking the wrong questions about AI and work — And it shows

Boards everywhere are asking about AI. That is the good news. The problem is that many are still asking the wrong questions.

The dominant boardroom conversation remains narrowly framed around adoption, efficiency, productivity, and cost reduction. How quickly can AI be deployed? How much labour can be optimised? What savings can be realised? What competitive advantage can be gained? Those are operational questions.

What is missing are the governance questions that determine whether the organisation remains resilient once AI becomes embedded into how work, decisions, and accountability actually function.

This distinction matters because AI is no longer simply a technology capability. It is rapidly becoming part of the enterprise operating model itself. It shapes how decisions are made, how customers are treated, how employees are managed, and, increasingly, how value is created and distributed.

Yet many boards continue to govern AI as though it were another digital transformation initiative. That framing is now dangerously insufficient. Organisations are moving aggressively to automate work because the economic incentives are compelling. AI can compress timelines, reduce labour intensity, and improve throughput at scale. Some organisations are already reporting measurable productivity gains and cost efficiencies.

But efficiency is not the same as resilience. A company can reduce costs while simultaneously weakening its ability to adapt, learn, govern, and defend its decisions under pressure. Many boards are not interrogating that tradeoff deeply enough.

The first wrong question is, “What work can we remove?”
The better question is, “What judgement must remain human?”

There are decisions where human involvement is not merely a legacy inefficiency but a governance safeguard. Contextual interpretation, ethical escalation, reputational sensitivity, and decision accountability are not problems to eliminate. They are controls. When organisations remove human oversight without clearly defining where authority, intervention, and challenge must remain, they create systems optimised for routine conditions but vulnerable when reality deviates from training assumptions. And reality always deviates.

The second wrong question is, “How much productivity are we gaining?”
The better question is, what capability are we eroding in the process?

Many AI efficiency gains come from compressing layers of work that historically developed future expertise—analysis, review, coordination, and operational judgement. Organisations may improve margins in the short term while quietly dismantling the pipelines that produce experienced managers, risk leaders, and institutional knowledge over time. This is not simply a workforce issue. It is a strategic continuity issue.

Boards that focus only on immediate gains risk overseeing organisations that become increasingly dependent on systems they do not fully understand and workforces they no longer know how to rebuild.

The third wrong question is, “Who owns the AI programme?”
The better question is, can accountability still be traced when decisions are distributed across systems?

Traditional governance assumes clear lines of responsibility. AI disrupts those assumptions. Decisions increasingly move across models, workflows, data pipelines, vendors, and operational systems. When outcomes fail, accountability becomes difficult to locate and even harder to defend.

Yet many organisations still approach accountability as a naming exercise rather than a control discipline. Boards should be demanding evidence that accountability is demonstrable, traceable, and operationalised, not simply assigned in governance documents.

The fourth wrong question is, “Is the system accurate?”
The better question is, “What happens when the system encounters conditions it has never seen before?”

Evidence increasingly shows that AI systems perform strongly within defined boundaries but degrade outside them. Edge cases, unfamiliar environments, and novel events expose weaknesses that routine testing rarely captures. Without escalation pathways, rollback mechanisms, and human intervention standards, organisations risk scaling failure at machine speed. This is where governance is actually tested, not during normal operations but when systems encounter the unexpected. The posture required from boards must now change.

AI-driven workforce transformation cannot be governed primarily as a technology initiative or a cost optimisation programme. It must be governed as a structural redesign of how the enterprise makes decisions, develops capability, allocates accountability, and sustains resilience over time. That requires sharper interrogation, stronger control expectations, and greater discipline in defining where human authority must remain non-negotiable.

Organisations most exposed in the next phase of AI adoption will not necessarily be those that moved too slowly. They will be those who optimised aggressively without fully understanding what they were truly redesigning beneath the surface. And by the time the consequences become visible, the capability to correct them may already be gone.

Amaka Ibeji, Founder of DPO Africa Network, is a Boardroom Qualified Technology Expert and Digital Trust Visionary. She advises boards, regulators, and organizations on privacy, AI governance, and data trust, while coaching and fostering leadership across industries. Connect: LinkedIn amakai | [email protected]