In 2024, the IMF warned that AI could affect almost 40% of jobs globally, with advanced economies facing even greater exposure as AI reshapes both routine and high-skilled work. The response in many boardrooms has been immediate: accelerate adoption, improve efficiency, reduce labour intensity, and redesign workflows around automation.
Those pressures are real—AI is already reshaping software engineering, customer operations, financial services, logistics, legal review, cybersecurity, and enterprise decision-making. Organisations that ignore this shift risk becoming structurally uncompetitive. But there is a growing problem emerging beneath the productivity narrative—one that many organisations are not yet confronting directly.
Many organisations are redesigning work faster than they understand what makes them resilient in the first place. This is where boards must now intervene differently and decisively.
The dominant assumption in many AI strategies is that human labour is the primary inefficiency to optimise. Yet real-world deployment is increasingly exposing the limits of that thinking. Microsoft reportedly scaled back parts of its AI coding rollout after costs associated with heavy AI usage became difficult to sustain. Uber reportedly exhausted its 2026 AI coding tools budget within four months. What initially appeared to be straightforward labour substitution is proving operationally, financially, and architecturally more complex. This matters because AI economics do not behave like traditional software economics.
