AI promises a productivity windfall – who will reap the benefits? | Computer Weekly

TEXT START: Two previous industrial revolutions suggest gains arrive late and flow to capital first.

The Dissection

The article correctly identifies the historical pattern: productivity gains arrive after organisational upheaval, flow to capital before labour, and can reduce labour’s share of income. It also names the mechanism clearly: collective knowledge is embedded in privately controlled machinery.

Then it retreats. “Augment or automate” and “shared or concentrated wealth” are presented as policy choices, converting a structural conflict into a management menu. The article diagnoses capital capture, then refuses to follow the logic to its terminal consequence: once AI performs economically necessary work, labour loses bargaining power regardless of how politely the transition is designed.

The Core Fallacy

The core fallacy is treating competitive outcomes as discretionary choices. A firm may choose augmentation temporarily, but if automation delivers lower costs or superior performance, competitors force adoption. Augmentation is often only the staging ground: once expert tasks are formalised, they become easier to automate.

The article also assumes that because AI contains accumulated social knowledge, society therefore has a practical claim to own it. That is a moral argument, not a mechanism of control. Compute, models, data centres, energy, logistics, patents and distribution remain controlled by capital unless power is forcibly transferred.

Under the hardened DT framework, P1 makes replacement economically dominant, P2 prevents institutions from preserving stable human-only domains, and P3 removes the majority’s productive necessity. The J-curve may delay the impact. It does not reverse it.

Hidden Assumptions

• Firms can coordinate around augmentation even when automation creates a competitive advantage.
• Human expertise will remain scarce after AI destroys the entry-level pipeline and progressively absorbs higher-order skills.
• “So-so automation” will remain technically poor enough to preserve displaced workers’ leverage.
• Historical lags provide enough time for institutions to adapt rather than merely postpone the shock.
• Governments can redistribute AI rents before ownership and control become entrenched.
• UBI or public services can substitute for lost productive participation without creating a permanently dependent population.
• Recognition that knowledge is collective will somehow produce collective ownership.
• Productivity gains will eventually translate into wages, despite the article’s own evidence that they need not.

Social Function

This is a partial-truth transition-management text wrapped in ideological anesthetic. It is not pure copium: it honestly records capital-first gains, labour-share decline, deskilling and displacement. But its insistence that the outcome is mainly a “choice” gives institutions a comforting escape hatch. It allows suppliers, governments and professional commentators to discuss humane deployment without confronting ownership, coercion and the collapse of labour’s bargaining position.

Its UBI and public-utility direction also changes the question from “Will people remain economically necessary?” to “Can people continue consuming?” That preserves demand. It does not preserve productive citizenship.

The Verdict

The article sees the first half of the autopsy and stops before the corpse is cold. It correctly predicts that AI’s early gains will flow upstream to capital and that productivity will not automatically raise wages. It fails by treating augmentation, redistribution and collectivisation as available choices within a system driven by competitive compulsion.

The likely outcome under DT is not a universally shared windfall. It is AI-owned productive capacity, a shrinking wage base, and transfers used to keep displaced populations consuming. Unless control of AI capital changes before P1–P3 harden, the article’s “fork in the road” is mostly theatre: capital takes the productive gains, while labour receives managed obsolescence.