Cognitive automation may arrive in employment statistics less as a wave of dismissals than as a vacancy that disappears. Firms can reduce demand for labour through hiring freezes, contractor cuts, non-replacement and attrition long before they announce that a profession has been automated.
Replacement begins without a dismissal
A company rarely starts a technological transition by dismissing every employee whose tasks can be automated. Contractors leave. Agency budgets shrink. Fixed-term roles expire. Retirees are not replaced. Teams discover that they can postpone recruiting another analyst, junior developer or paralegal. The first person replaced by a capable model may be someone who was never hired.
Current AI is much weaker than the hypothetical superhuman system considered in the broader economic argument, yet the OECD already finds that recent advances expose high-skilled cognitive work in ways earlier automation did not.1 More capable systems would intensify the pressure on functions built around standard digital inputs and outputs: document review, routine coding, translation, reporting, tax preparation, procurement analysis, customer support, scheduling and compliance administration.
A department of one hundred people might not become a control room of twelve overnight. It can still move in that direction each time a position becomes vacant. The survivors would not necessarily be those with the highest exam scores. Process owners, accountable signatories, relationship holders and experts in local exceptions may remain valuable after much of the underlying analysis is machine-made.
Germany separates insiders from entrants
The route matters particularly in Germany. Established firms cannot usually dismiss protected employees with a casual announcement that software now performs their tasks. The Kündigungsschutzgesetz requires qualifying dismissals to be socially justified. Operational dismissals depend on urgent business requirements, possible continued employment and social-selection criteria.2
Those protections create time and bargaining space. They do not require an employer to preserve the flow of new entrants. A firm can keep current staff while stopping graduate recruitment, reducing contractors, allowing temporary contracts to expire, moving new work into a subsidiary or waiting for older employees to retire.
Germany’s demography makes this path plausible. Destatis projects that one quarter of the population will be at least 67 years old by 2035; the number of people of working age is projected to decline over the longer term.3 A shrinking workforce can cushion the aggregate unemployment effect while making automation through non-replacement easier.
The protected 55-year-old may keep a position whose substance has been hollowed out. The 24-year-old never enters the profession. Headline employment can remain comparatively stable while the distribution of opportunity changes sharply between insiders and outsiders.
Transition instruments assume there is a transition to complete
Germany already has instruments for temporary disruption and structural adjustment. The federal labour ministry lists short-time work, training support, qualification allowances and transfer benefits among its labour-market tools; it describes Kurzarbeitergeld as support for a temporary loss of work.4
These tools can preserve employment through a cyclical fall in demand or finance movement into a growing occupation. They are less complete answers when the change is a durable reduction in demand for a broad class of cognitive work. A bridge for a temporary shock cannot serve as a permanent employment model.
Retraining also presumes that a reasonably accessible destination exists. If several adjacent office professions compress at the same time, moving people between them does not solve the structural problem. Training can still help individuals and address shortages elsewhere, especially in work constrained by physical capacity. It should not be described as evidence that the total demand for entry-level cognitive labour will recover.
The missing opportunity is easy to miss
Dismissal figures measure a visible event. They do not show the graduate intake that was cancelled, the contractor budget that did not return or the role absorbed by a team using better tools. Policy built around reported redundancies will see the change late.
The first useful indicators are therefore flow measures: vacancies per output, entry-level hiring by function, conversion of temporary roles, replacement rates after retirement and the share of work shifted to contractors or subsidiaries. These measures distinguish a stable workforce from a stable path into that workforce.
Closing entry routes also damages the way professions create experienced practitioners. That apprenticeship problem deserves its own treatment; it is not necessary to resolve it to see the immediate employment consequence.
Labour protection can make technological change less arbitrary for people already inside an institution. It cannot, by itself, give outsiders access to a profession that has stopped hiring. Cognitive automation may therefore produce a divided market: protected incumbents retain positions for years while a cohort of would-be entrants discovers that the first rung has been removed.
Footnotes
-
“Artificial intelligence and the labour market: Introduction”, OECD Employment Outlook 2023. ↩
-
Kündigungsschutzgesetz § 1, Federal Ministry of Justice and Federal Office of Justice. ↩
-
16th coordinated population projection, Destatis, December 2025. ↩
-
Instrumente der Arbeitsmarktpolitik, Federal Ministry of Labour and Social Affairs, December 2025. ↩