The Insight
What's Really Happening
The assumption underpinning many AI strategies is simple: progress follows expertise. If you can recruit enough machine-learning engineers, data scientists, or AI leaders, transformation will follow. Yet evidence from across industries now shows the opposite pattern. Teams grow. Costs rise. Outcomes plateau.
A recent synthesis of executive research makes the point starkly. A Boston Consulting Group study found that around half of organisations experimenting with AI have failed to scale it meaningfully, remaining stuck in pilots despite heavy investment. Fortune, surveying thousands of executives, reported that nearly 90 per cent saw little or no productivity impact from AI to date. These are not laggards. They are committed adopters.
The explanation is not a shortage of intelligence. It is a surplus of friction.
AI capability is often treated as an accumulation of individual skill. In practice, it behaves more like an emergent property of systems. Exceptional hires can sense opportunities, but they cannot seize or embed them without authority, data access, and integration into real workflows. Where those conditions are absent, talent becomes stranded.
The research is consistent. Organisations build centres of excellence, recruit elite specialists, and expect velocity. What they create instead are bottlenecks. Business units wait for models they cannot deploy. Governance teams intervene late. Data remains fragmented. Decision rights are unclear. AI outputs become advisory artefacts rather than operational levers.
One widely cited statistic captures the frustration: in 2025, nearly half of enterprise AI proofs of concept were abandoned before reaching production. Not because the models failed, but because the organisations around them could not absorb them.
This is what many practitioners now call capability theatre: headcount rising, value flatlining.
The Strategic Shift
Why It Matters for Business
The implication for leaders is uncomfortable but clear. AI progress does not compound through recruitment alone. It compounds through structure.
AI systems change how decisions are made, how work flows, and how risk is distributed. That places them firmly beyond the remit of IT or HR alone. They cut across operating models, governance frameworks, and incentive structures. When those remain unchanged, even world-class talent hits a ceiling.
Consider the distinction between talent and capability. Talent is portable. It can be hired, poached, or lost. Capability is embedded. It lives in workflows, platforms, and decision rights. It persists even when individuals leave.
High-performing organisations are beginning to design for this reality. Instead of asking, “Who should we hire?”, they ask different questions:
Who owns this AI system two years after launch? Where does the authority to act on its outputs sit? How quickly can insights move from model to decision to action?
This shift reframes AI from a project to a system, from something delivered to something that evolves. It also exposes why traditional organisational models struggle. Centralised approval structures slow learning. Fragmented data architectures starve models of context. Incentives that reward caution over iteration discourage adoption.
Deloitte's research into AI ROI highlights the pattern. Teams that report strong returns are significantly more likely to be cross-functional, empowered to make decisions, and embedded in business units rather than isolated in technical silos. The technology is similar. The outcomes are not.
The strategic risk is not merely inefficiency. It is churn. Elite AI hires increasingly scrutinise authority, autonomy, and mission clarity before accepting roles. When they encounter environments where insight cannot translate into action, they leave. The organisation is left with higher costs and less institutional memory.
The Human Dimension
Reframing the Relationship
From the inside, the experience is corrosive. You hire a senior specialist to “lead AI”. They quickly discover that data access requires negotiation, deployment requires escalation, and risk sign-off arrives too late to matter. Their work becomes explanatory rather than transformative.
From the outside, the signals are just as clear. Candidates now ask sharper questions: Who decides? What can I change? How will success be measured? Compensation matters, but structure matters more. Even billion-dollar offers have been declined when purpose and autonomy were absent.
For the wider workforce, the consequences are subtler. AI tools surface insights faster than organisations can absorb them. Employees learn that acting on AI-driven recommendations carries personal risk, while ignoring them carries none. Over time, trust erodes. Systems are bypassed. Shadow tools emerge. The organisation appears modern but behaves conservatively.
You cannot fix this with another hire.
The Takeaway
What Happens Next
By 2026, the advantage will belong to organisations that treat AI capability as a design challenge, not a recruitment race. Hiring remains necessary, but it is no longer sufficient.
The leaders who progress will be those who redesign how decisions are owned, how data is shared, and how learning is rewarded. They will embed AI into workflows, grant authority alongside insight, and fund systems that evolve rather than projects that end.
The enduring lesson is simple and uncomfortable: Talent amplifies structure. It does not replace it.
