Every week, another warning claims that artificial intelligence is poised to erase white-collar jobs, with coders, analysts and office workers supposedly months away from being replaced by systems that work faster and without fatigue. Markets have reacted. Corporate executives have drawn up contingency plans.
But a February 2026 report from Citadel Securities, titled The 2026 Global Intelligence Crisis, argues that the data does not support the urgency of those fears. After reviewing labour market indicators, AI adoption surveys, computing economics and historical patterns of technological change, the firm concludes that the widely discussed displacement crisis remains largely theoretical.
Software engineer job postings have rebounded. Daily AI use for work has not accelerated. And the economics of large-scale compute, Citadel Securities argues, impose real limits on how quickly human labour can be substituted.
Software engineer hiring is recovering despite the narrative
One of the clearest data points in the Citadel Securities report is the recovery in software engineer job postings on Indeed. After declining from early 2024 into mid-2025, postings have risen approximately 11 percent from their trough by January 2026, based on the firm’s tracking of a 21-day moving average. Overall Indeed listings have also edged higher, though less sharply.
If AI coding tools were displacing engineers at scale, demand for those engineers would likely be falling, not recovering.
“The data seems unexpectedly stable and presents little evidence of any imminent displacement risk,” wrote Citadel Securities.
The firm adds that its broader labour market tracker, which aggregates signals from ADP, ISM, Revelio, Indeed and private payroll measures, continues to point towards improvement. At the same time, AI-driven data centre construction has supported hiring in construction and engineering roles tied to infrastructure build-out.
Daily AI use for work has not accelerated
Citadel Securities places significant weight on how intensively AI tools are being used, rather than whether workers have merely tried them. Drawing on the St Louis Federal Reserve’s Real Time Population Survey, the firm separates occasional use from daily work integration.
The share of working-age adults using generative AI every day for work has remained roughly between 10 and 12 percent from August 2024 through November 2025, according to the report. Broader measures of “any use” have risen, but intensive daily usage has not shown meaningful acceleration.
“The first order presentation of AI adoption is generally a binary question: Do you use AI? The more important question insofar as it relates to the AI displacement narrative is: how intensely is AI being used for work?” wrote Citadel Securities.
The firm argues that occasional assistance from AI does not equate to full automation of a role. In its reading of the data, the feared inflection point in labour substitution is not yet visible.
Technology adoption has historically followed an S-curve
Citadel Securities challenges the assumption that because AI systems improve rapidly, economic adoption must accelerate at the same pace. The report distinguishes between technological capability and economic deployment.
“Technological diffusion has historically followed an S-curve. Early adoption is slow and expensive. Growth accelerates as costs fall, and complementary infrastructure develops. Eventually, saturation sets in, and the marginal adopter is less productive or less profitable which causes growth to decelerate,” wrote Citadel Securities.
Generative AI adoption, the firm notes, is tracking along a pattern that resembles prior technologies. After roughly three years of availability, generative AI has reached about 40 percent adoption among working-age adults in some capacity, comparable to the early trajectory of personal computers and still below the internet at a similar stage.
“Recursive capability does not imply recursive adoption,” the report states.
Compute economics create a natural boundary
The report argues that replacing large segments of white-collar labour would require computing capacity far beyond current utilisation levels.
“Displacing white collar work would require orders of magnitude more compute intensity than the current level utilization. If the marginal cost of compute rises above the marginal cost of human labor for certain tasks, substitution will not occur, creating a natural economic boundary,” wrote Citadel Securities.
Training and running advanced models requires substantial semiconductor capacity, energy and data centre infrastructure. The firm argues that economic deployment remains bounded by physical capital, energy availability, regulatory approvals and organisational change. Improvements in model capability do not automatically make mass substitution economically rational.
Productivity gains do not automatically destroy demand
Citadel Securities also disputes the idea that higher productivity necessarily leads to fewer jobs overall. The firm characterises AI as a productivity shock, which lowers the cost of producing goods and services.
“Lower prices increase real purchasing power, which generally increases consumption. A scenario in which productivity surges but aggregate demand collapses while measured output rises violates accounting identities,” wrote Citadel Securities.
The report points to strong US new business formation as evidence that economic dynamism remains intact. Historically, the firm notes, technological revolutions have altered the composition of work rather than eliminated labour as an input.
“It seems more likely that AI will be a complement rather than a substitute for labor in many areas. Historically, technological revolutions have altered task composition rather than eliminated labor as an input,” wrote Citadel Securities.
Keynes and the elasticity of human wants
The report references John Maynard Keynes, who predicted in 1930 that productivity gains would eventually reduce the workweek dramatically. That forecast did not materialise, not because productivity failed to rise, but because consumption expanded.
“Keynes underestimated the elasticity of human wants,” wrote Citadel Securities.
Rising productivity has historically funded new forms of consumption and new industries rather than permanent labour contraction.
What would it take for AI to cause a demand shock?
Citadel Securities does not dismiss the possibility of disruption. Instead, it lays out the conditions that would need to occur simultaneously for AI to generate a sustained macroeconomic demand shock.
“For AI to produce a sustained negative demand shock, the economy must see a material acceleration in adoption, experience near-total labor substitution, no fiscal response, negligible investment absorption, and unconstrained scaling of compute,” wrote Citadel Securities.
Based on current data, the firm concludes those conditions are not in place. Software hiring has stabilised and improved, daily work usage of AI remains limited, compute constraints persist, and institutional responses would likely slow large-scale displacement.
The firm’s central argument is measured rather than dismissive. AI is powerful, but its economic diffusion is constrained. The evidence so far suggests adjustment and complementarity, not imminent collapse, it explained.