Train workers to use AI or risk 'productivity doom loop', warns Euan Blair - The Observer

TEXT START: The chief executive of the UK’s largest apprenticeships provider has warned of the hit to productivity caused by workers never having been “shown how to do their jobs differently”.

The Dissection

This is a commercial transition-management pitch disguised as a warning. The article correctly identifies a real bottleneck: workers who cannot integrate AI will waste time and fall behind. But it reframes structural labor displacement as a training deficit. That conveniently turns mass anxiety into revenue for retraining vendors and responsibility for workers.

Multiverse’s own position exposes the mechanism. It sells AI adaptation while using AI to monitor and rate teachers, producing the “horrendous stress” described in the article. The promised solution to insecurity is deeper integration into the system generating the insecurity.

The Core Fallacy

The article assumes that making workers more productive makes them safer. Under the Discontinuity Thesis, the opposite is often true. AI-skilled workers help firms produce the same output with fewer people. Training can therefore accelerate substitution while improving the bargaining position of only a narrow layer of survivors.

The “productivity doom loop” is real, but Blair misidentifies its endpoint. Poor adoption may damage competitiveness and trigger layoffs. Better adoption may also trigger layoffs by making labor less necessary. The article treats productivity as a job-preservation mechanism when, under P1–P3, it is the weapon that severs employment from production.

Hidden Assumptions

• Productivity gains will create enough new work to absorb displaced workers.
• Employers will share efficiency gains rather than capture them as profit, cost reduction, or headcount reduction.
• AI will remain mainly complementary instead of becoming superior across cognitive tasks.
• “AI literacy” produces durable indispensability rather than temporary usefulness.
• Union protection can be secured through skills training instead of control over capital, deployment, and surplus.
• Better worker performance data will empower workers rather than enable surveillance, ranking, and automated elimination.
• The £50bn fiscal benefit cited by the OBR will translate into worker security rather than state and capital gains.

None of these assumptions is established by the productivity figures presented. Measuring inefficient data time does not demonstrate that the resulting efficiency will preserve employment.

Social Function

This is transition management, elite self-exoneration, and ideological anesthetic—with a substantial partial truth. It tells workers and unions that survival depends on becoming more useful to the machinery replacing them. It shifts the burden from owners and controllers of AI capital onto employees who must continuously retrain under threat of exclusion.

It also normalizes AI surveillance as “customer experience” and presents labor discipline as protection. Training is the respectable face of selection: a few workers become Servitors, while the rest become cheaper to replace.

The Verdict

The article sees the fracture but misnames the disease. AI training is valuable for individual survival and essential for firms seeking transition advantage. It is not a solution to mass employment collapse. Under the Discontinuity Thesis, upskilling is more likely to accelerate the destruction of routine cognitive labor than prevent it.

The real exit is not universal training. It is ownership or control of AI capital, energy, logistics, maintenance, verification, or transition chokepoints. For everyone else, Blair’s prescription is not rescue. It is a better-prepared march toward redundancy.