Agentic-AI Reshapes Labor Demand in Finance
The paper "From Clerks to Agentic-AI" by Lu Yu and Xiang Li documents how successive technology waves altered labor scale in asset management. Using a small panel of representative firms the authors track a simple productivity measure, assets under management per employee, and compare it to revenue per employee and operating expense intensity across three technological waves: computerization (1980s-1990s), indexing and passive investing (2000s-2010s), and the AI and automation wave (2015-present). The study does not claim causal identification; it provides stylized facts on scale effects and operating leverage in asset management as technology shifts work from clerical processing toward higher-value supervision and strategy.
What happened
The paper From Clerks to Agentic-AI by Lu Yu and Xiang Li documents how technology changed the labor scale in finance by measuring assets under management per employee across three historical waves: 1980s and 1990s (computerization), 2000s and 2010s (indexing and passive investing), and 2015 to the present (AI and automation). The authors assemble a small panel of representative asset managers and report stylized facts on productivity and expense structure rather than causal estimates.
Technical details
The analysis centers on three firm-level metrics tracked over time:
- •assets under management per employee
- •revenue per employee
- •operating expense intensity
The paper compares trends in these measures to reveal scale effects and shifting operating leverage as firms adopt new technologies. The empirical strategy is descriptive; the authors explicitly avoid causal identification. Data construction, firm selection, and panel composition are documented in the manuscript and the PDF available on arXiv.
Context and significance
This work situates current AI-driven automation in a long-run history of productivity shifts in finance. The sequence from clerks to computerized back offices to passive strategies has been associated with changing labor requirements per AUM; the ongoing AI wave is framed by the authors as the latest phase with the potential to further automate various asset management tasks. For practitioners, the paper provides quantitative anchors to reason about likely displacement and complementarities: changes in AUM per employee could imply both headcount compression in routine roles and rising demand for oversight, model governance, and product design skills when automation is deployed.
Limitations and caveats
The study is descriptive and based on a small panel; results are stylized facts, not causal claims. Heterogeneity across firm business models, regulatory regimes, and timing of AI adoption limits direct extrapolation to every firm or subsegment.
What to watch
Follow-up work should expand the sample, link adoption timing to wage and occupation-level outcomes, and measure how agentic-AI systems change task composition versus simple automation. For firms, key signals will be sustained changes in AUM per employee, increased attention to model governance, and reallocation of headcount toward oversight and product roles.
Scoring Rationale
This arXiv paper provides useful empirical documentation of technology-driven scale effects in asset management, offering practitioners quantitative anchors. It is descriptive and based on a small panel, so it is informative rather than paradigm-changing.
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Ad Tech problemsStep-by-step roadmaps from zero to job-ready — curated courses, salary data, and the exact learning order that gets you hired.
