The correction is real. The attribution is wrong.

Two of the most consequential voices in the global technology industry have recently pushed back against the dominant narrative around AI and job losses. Demis Hassabis, CEO of Google DeepMind, questioned whether AI is suddenly replacing large portions of the workforce. Jensen Huang, CEO of NVIDIA, called the AI-layoff narrative lazy and irresponsible. Both are correct. Both stop short of the deeper structural argument. That gap matters enormously – because a misdiagnosis at this scale produces the wrong policy response, the wrong investment thesis, and the wrong organisational strategy.

The correction underway across the technology sector is real. Thousands of roles are being removed at Meta, LinkedIn, Wix, Webflow, and across the broader tech industry – the visible 2026 count already exceeds 142,000. But the causal story being offered – that companies over-hired in 2020 and 2021 and are now correcting back to rational headcount, with AI playing no structural role – is factually incomplete. It conflates two distinct phenomena: a cyclical inventory correction, and a structural change in how value is created and where employment is generated in the economy. Treating both as the same event produces the wrong response at every level.

Bloor Research modelling of ONS payroll and HMRC data identifies a structural employment gap of approximately 1.3 million roles in the UK alone – positions that historical GDP-to-employment ratios would predict to exist, but which have not materialised. Our methodology excludes the 2020 COVID distortion from the structural gap calculation to protect the argument from cyclical dismissal. The same pattern holds across the EU, where UK and European job creation rates track near-identically across the same period, isolating technology and offshoring as the structural variables rather than any country-specific policy factor.

The mechanism is Ghost GDP: economic output that grows without generating proportional human employment. Digital workers, automated systems, and AI capability now contribute to productive output alongside human workers. GDP can expand, revenue can grow, and value can be created – but the human headcount required to produce that value is no longer what historical models predict. This decoupling has been present in the data for more than two decades. AI has accelerated a condition that was already structurally embedded.

The Hierarchy tax and the asset-liability misclassification

The second mechanism operating beneath the current correction is the Hierarchy Tax: the measurable cost of management layers that exist primarily to coordinate and route information rather than to create value directly. When AI systems perform coordination at negligible marginal cost, those layers become visible as structural liabilities rather than operational necessities. The headcount being removed in current efficiency announcements is, in many cases, the Hierarchy Tax crystallising under AI load.

This is the asset-liability misclassification that sits at the root of most current workforce strategy errors. In standard accounting, employees appear on the balance sheet as human capital — an asset. In economic reality, fixed-cost human employment in a dynamic AI environment behaves as a liability: a committed future obligation whose productive output is increasingly matched or exceeded by digital alternatives at lower marginal cost. The organisation that does not distinguish between the human roles that appreciate in value as AI capability increases – because they govern, design, and architect the hybrid system – and those that depreciate because they execute tasks now within AI’s range, is not managing its balance sheet. It is managing a headcount number while the underlying asset-liability structure changes beneath it.

Bloor Research observes, across organisations with advanced AI deployment, layoff concentrations of 26 to 65 percent in higher-rate salary brackets – roles in the £50,000 to £125,000-plus range. These are structural removals of knowledge-intensive positions that will not return in the same form. The knowledge carried in those roles is being extracted into digital systems or lost entirely, representing Labour Debt transferred onto a reduced human workforce. The Ghost Liability – the contingent exposure in the remaining workforce and the undisclosed productivity assumption now embedded in AI systems – does not appear on any current balance sheet format.

Removing headcount without redesigning the operating model that generated the Hierarchy Tax does not resolve the structural problem. It reduces the immediate cost. The architecture that produced it remains intact and will regenerate it in a new form.

The correction is real. The attribution is incomplete. Until organisations name the structural cause accurately – Ghost GDP, the Hierarchy Tax, the asset-liability misclassification – the responses they design will address the symptom and compound the condition.

Print Article for this blog Download here