AI and Digital Darwinism: Survival of the Fastest.

Why the productivity dividend is real, uneven, and already separating winners from casualties

For investors and boards, the question is no longer whether AI works. It does. The real question is who can convert disruption into durable advantage and who will see their competitive position erode.

Artificial intelligence is delivering a real productivity dividend, but not evenly. In aggregate, the medium-term outlook for AI and jobs is still net positive. In practice, the transition is already proving painful for specific cohorts: white-collar entry-level workers, new graduates, and regions whose labour pools do not match the capabilities that AI is increasing in value.

Whether firms convert disruption into durable advantage depends less on simple adoption than on redesign of workflows, workforce models, and operating discipline. AI is no longer just a technology theme. It is becoming a selection mechanism, with human capital at the centre.

Productivity appears first as divergence

The classic pattern of every general-purpose technology is that productivity does not show up immediately. It appears only after organisations change how work is actually done. AI is following the same path. Adoption is broad, but deep implementation remains rare — which is why widespread use still coexists with modest P&L impact at the median firm.

The most important point is that productivity is showing up first as dispersion. Leaders are pulling away from laggards. PwC's 2026 Global AI Jobs Barometer found that the top 20% of the most AI-exposed companies achieved average labour productivity growth of 163% relative to their 2018 baseline — nearly five times higher than less-exposed peers. That is not a story of universal uplift. It is a story of widening separation.

The first labour shock is in hiring

The labour-market impact of AI is not yet primarily a layoff event — it is a hiring one. Entry-level hiring has already fallen 16% in AI-exposed functions (Brynjolfsson, 2025). The earliest signs show up in vacancy data, graduate recruiting, junior-role redesign, and non-replacement of leavers. By the time unemployment rates move visibly, several years of structural change may already have been absorbed beneath the surface.

This distinction matters. If a firm no longer hires analysts, assistants, junior coders, or operations trainees at the same pace as before, the labour market weakens without immediately showing large-scale dismissals. That creates a ladder-loss effect: the first rung of the career ladder disappears just as workers most need access to it.

This phase should therefore be understood as a mobility shock more than a classic employment shock. Senior workers often benefit first because AI amplifies judgment and synthesis. Entry-level workers are more exposed because many apprenticeship tasks are exactly the tasks AI now handles well. Understanding the backlash this mobility shock will create is important.

The net-positive story hides transition risk

Forecasts project 170 million jobs created and 92 million displaced globally by 2030 — a net gain of +78 million (WEF Future of Jobs Report, 2025). That is directionally encouraging, but it should not be mistaken for a transition plan.

The problem is asymmetry. 77% of new AI jobs are expected to require master's degrees. A displaced bank teller cannot retrain as a Big Data Specialist in 12 months. The jobs being created are often not in the same locations, sectors, or skill bands as the jobs being displaced. The speed of labour conversion is therefore more important than the headline arithmetic.

Policymakers and employers should focus less on whether AI is net positive in theory and more on where frictions are building: graduate pipelines, retraining quality, geographic mismatch, and the shrinking availability of developmental early-career work.

Why most firms still do not capture value

A potential strategic error is to confuse use with transformation. Giving employees access to AI tools is not the same as redesigning the firm around them. The former improves local efficiency. The latter impacts competitive position in a fast-changing world.

The scale of the gap is striking. McKinsey’s State of AI in 2025 finds that while 88% of organisations use AI in at least one function, only around one‑third have scaled it across the enterprise, and just a small minority qualify as ‘AI high performers’ generating material EBIT impact from AI. That minority is already resetting the cost and capability floor for entire sectors

Three capability gaps explain why most firms remain stuck:

– AI Fluency: can people use the tools competently and integrate them into daily work? Necessary, but no longer a moat.

– Innovation: can the firm recombine AI outputs into new propositions and client value rather than merely cheaper execution?

– Implementation: can management embed AI into workflows, controls, and governance strongly enough for the gains to persist and scale? Only 18% of firms have moved AI beyond pilot into core operations.

Once AI becomes broadly accessible, advantage shifts away from tool access alone and toward the organisational capacity to direct, govern, and underwrite its outputs. The strategic issue is no longer whether AI can raise output but which firms can capture that output as margin, market share, or strategic control.

The real premium is judgment

In this environment, fluency becomes the entry ticket. Judgment then becomes the premium. As AI makes information processing cheaper and faster, the scarce capability shifts upward: problem framing, accountability, prioritisation, ethical restraint, interpretation, and trust.

This is the basis of the synthesis economy. Machine intelligence expands the supply of analysis, drafting, coding, and structured output. Human advantage increasingly lies in deciding what matters, what is credible, what is commercially relevant, and what should actually be done.

The firms that win will not be those that automate the most blindly. They will be those that combine machine execution with human underwriting — a disciplined human-machine stack in which intent, control, and responsibility remain clear.

A practical Digital Darwinism selection framework

A useful way to assess winners and losers is through a simple 2×2: AI adoption speed on one axis, human capital quality on the other.

– Fast adopters with strong human capital are the likely durable winners. They set the new cost floor and the new capability floor.

– Slow adopters with strong human capital still have value, but their window is limited. Expertise without adoption decays within a 2–3 year window.

– Weak adopters with weak human capital become stranded. They compete on neither cost nor quality with weak human capital may look impressive briefly, but they are fragile — lower costs in the short term, higher operational and governance risk in the medium term.

That framework applies not only to firms, but also to workers and countries. The same logic explains why AI is likely to widen gaps in wages, margins, and national competitiveness unless adaptation speeds up materially.

What matters now

AI is not one shock. It is a multi-speed sorting process. The indicators worth watching are not just unemployment or technology spending — they are hiring patterns, junior-role attrition, workflow redesign, training depth, governance quality, and the speed with which firms move from pilots to core operations.

The broad direction is clear. AI will create value, but it will not be shared evenly or captured automatically. The winners will be the organisations that treat adaptation as a core strategic capability rather than a software deployment exercise. In that sense, Digital Darwinism is not a metaphor. It is becoming a market reality.

Sources: PwC, Global AI Jobs Barometer 2026, Brynjolfsson et al., 2025, World Economic Forum, Future of Jobs Report 2025, McKinsey & Company, State of AI in 2025 (Global AI Survey), McKinsey / Stanford HAI / Penn Wharton, 2024–2026

© Maisonneuve Global Advisors Ltd. 2026. For information and discussion purposes only. This article does not constitute investment advice.