New report claims AI is helping create jobs in UK businesses, as workers can't get enough of it

TEXT START: AI is actually creating more jobs than it's replacing

The Dissection

This is not evidence that AI creates durable mass employment. It is a vendor-backed optimism narrative built from perceptions, executive self-reporting, and vague definitions of “new roles.” The article converts AI adoption activity—implementation, oversight, training, integration, and management—into proof that the employment system is safe.

The headline also outruns the data. “Workers can’t get enough of it” is promotional language. The supplied figures show partial adoption, substantial fear of replacement, and sharply higher agent use among managers and IT decision-makers than ordinary knowledge workers. That is not worker emancipation. It is an ownership-and-control gradient.

The Core Fallacy

The central error is confusing transitional job creation with a stable replacement for productive participation.

AI can create roles while organizations are installing, supervising, selling, auditing, and adapting the technology. Those roles are complements to the expansion phase. They do not prove that AI will preserve the larger population’s bargaining power once the systems become cheaper, more capable, and operationally mature.

The survey also does not establish net employment. “Created new roles” is not compared against total layoffs, hiring freezes, reduced hours, wage compression, contractor elimination, or productivity-driven reductions elsewhere. Nor does it show whether the new roles are permanent, numerous enough to absorb displaced workers, or accessible to the same people being displaced.

Under the Discontinuity Thesis, this is a lag phenomenon: the machine is first surrounded by human scaffolding. The scaffolding is then mistaken for the building.

Hidden Assumptions

• That executives accurately report employment effects rather than marketing successful adoption.
• That “new roles” means durable, economically necessary jobs rather than temporary transition work.
• That gross job creation matters without measuring net headcount and wage effects.
• That workers who use AI remain indispensable after AI agents improve.
• That increased AI usage represents worker control rather than dependence on systems controlled by firms.
• That roles created in one organization or sector can absorb workers displaced across the economy.
• That current adoption patterns remain stable as AI agents become more autonomous.
• That the 10% reporting role reduction captures the full effect of substitution.
• That confidence from business leaders is stronger evidence than the 38% of knowledge workers already fearing replacement.
• That skill adaptation can outrun the declining price and rising capability of automated cognition.

Social Function

Primary classification: ideological anesthetic and transition management, with elements of corporate propaganda and elite self-exoneration.

The article gives firms a reassuring script: AI is not destroying the labor system; it is merely “changing work,” and workers who adapt will prosper. That framing relocates responsibility from structural displacement to individual skill acquisition. It also lets AI vendors present adoption as benevolent job creation while the control of the productive machinery remains concentrated above the ordinary worker.

There is a partial truth buried inside it. New transition roles can appear, and AI can temporarily complement human labor. But the article treats that temporary complementarity as a refutation of long-run substitution. It is not.

The Verdict

This report does not show that AI is saving employment. It shows that the early deployment phase still requires human operators, managers, and institutional scaffolding—and that those closest to organizational power are adopting the tools fastest.

The claim is structurally weak because it measures reported optimism and gross role creation, not durable mass participation. Under P1, the transition roles are exposed to the same automation they are helping deploy. Under P2, firms have no durable reason to preserve human-only cognitive domains when competitors can automate them. Under P3, the decisive question is whether the majority remain economically necessary after deployment matures. This article provides no evidence that they will.

The corpse is not alive because the undertakers are still being paid.