From Clerks to Agentic AI: How Will Technology Transform the Labor Market in Finance?
Lu Yu 1, Xiang Li 2 · 2026
Read From Clerks to Agentic AI: How Will Technology Transform the Labor Market in Finance? by Lu Yu 1, Xiang Li 2 (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-19833.txt.
Introduction
Yu and Li track how technology changed labor needs in finance across three waves. The waves are computerization in the 1980s and 1990s, indexing and passive investing in the 2000s and 2010s, and AI and automation from roughly 2015 to today. Using a small panel of representative firms, they compare assets under management per employee, revenue per employee, and operating expense intensity over time. The goal is not causal claims but a record of stylized facts about how technology changed the scale of asset management work.
Catalog record
| CREATOR | Lu Yu 1, Xiang Li 2 |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 19262 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.19833
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
From Clerks to Agentic AI: How Will Technology Transform the Labor Market in Finance? Lu Yu 1, Xiang Li 2 Abstract Artificial intelligence is beginning to reshape financial markets in ways that echo earlier technological transitions in finance. This paper examines how task automation changes labor productivity and labor demand in U.S. financial firms across three technology waves: computerization, indexing, and agentic AI. Combining historical comparison with firm-level operating outcomes and filing-based automation measures, it argues that successive automation waves reorganize workflows, scale, and the distribution of capability across firms rather than simply eliminating jobs. The resulting pattern is likely to be a more polarized industry in which large institutions retain structural advantages, small teams become more capable, and middle-layer functions face the greatest pressure. I. How Does Agentic AI Impact Financial Markets? Finance is an unusually informative setting for studying automation because it combines standardized workflows, information processing, client service, and judgment-intensive decision making within the same firms. New technology therefore affects tasks unevenly: some activities become cheaper and faster almost immediately, while others remain constrained by supervision, trust, interpretation, and accountability. The relevant question is not simply whether AI reduces headcount, but how automation changes the internal allocation of tasks and what that implies for productivity, scale, and labor demand. The analysis links historical technological change to contemporary AI adoption. Rather than treating AI as a purely speculative future shock, it examines whether rising automation intensity in firm disclosures is associated with changes in revenue per employee, assets under management per employee, and labor expense per employee. The broader aim is to place agentic AI within a longer sequence of technological change in finance rather than to treat it as a complete break from the past. The contribution is threefold. First, the paper organizes the analysis around three economically meaningful waves: computerization, indexing, and AI-driven automation. Second, it constructs filing-based measures of automation intensity that map onto these waves. Third, it relates those measures to firm-level operating outcomes to assess whether finance becomes more scalable and productive before labor-cost adjustment becomes visible. The preliminary evidence is consistent with that view. Productivity rises across successive technology waves, labor-cost adjustment appears slower than output adjustment, and AI-era firms exhibit the strongest filing-based automation intensity. These patterns are descriptive associations, not causal estimates, but they provide a disciplined framework for studying how automation may reorganize labor in finance. II. Historical Lessons from the Computerization and Indexing Eras The natural benchmark for the current AI wave is the earlier computer revolution in finance. In the 1980s and 1990s, personal computers, spreadsheets, Bloomberg terminals, Reuters systems, and electronic trading infrastructure transformed how financial firms processed information and managed risk. Those technologies did not eliminate finance; they changed who could operate effectively, how quickly information moved through the system, and which firms could scale their edge. The second major wave came with indexing and rules-based investing. Between roughly 2000 and 2015, ETFs, benchmark replication, passive allocation, and automated portfolio implementation reshaped large parts of asset management and market intermediation. This wave mattered not because it replaced all active judgment, but because it codified a large set of investment and rebalancing tasks that had previously been more labor intensive or discretionary. The third wave is the current AI and workflow-automation era, beginning around 2015 and accelerating sharply with generative and agentic systems. Unlike the earlier computerization wave, which mainly expanded computation and connectivity, the AI wave pushes further into cognition-adjacent tasks such as monitoring, summarization, drafting, triage, and workflow coordination. Unlike indexing, which standardized portfolio construction in a narrower domain, AI has the potential to affect a broader range of occupations inside financial firms. These eras are descriptive and interpretive rather than a stand-alone identification strategy. Their purpose is to provide a historically grounded framework for comparing how different forms of task automation alter productivity and labor demand over time. The task-level economic logic is illustrated by the following cost comparison, which shifts the discussion from infrastructure to the practical choice among domestic labor, offshore labor, and AI agents for repeatable workflows. Figure 1: Workforce transformation cost comparison III. How Does Automation Raise Labor Productivity? Automation can raise labor productivity through several channels: it can reduce routine manual work, lower coordination costs, accelerate workflow speed, and enable firms to operate at greater scale without proportionate increases in labor input. In finance, these effects may appear in back-office processing, reporting, risk monitoring, trade support, and parts of research production. Automation, however, need not generate immediate labor displacement. Technology can change the composition of work well before it changes headcount. Firms may reassign workers toward supervision, exception handling, and client-facing tasks while software absorbs repetitive tasks in the background. Labor expense per employee may therefore respond more slowly than output-based measures of productivity. The core conceptual variable is task_automation_intensity , defined as the extent to which routine, codifiable, or repetitive work is shifted from human labor to software, rules, machines, or AI systems. The vocabulary changes across eras, but the economic concept remains stable: firms differ in how intensively they automate tasks previously performed by people. This framework yields three testable implications. First, productivity should rise more in later technology eras than in the pre-computerization benchmark. Second, higher task-automation intensity should be associated with higher output per worker or greater scale per worker. Third, labor expense per worker may move more slowly than output measures if automation primarily works through reorganization and augmentation rather than immediate labor shedding. IV. How Do We Define Productivity and Automation Intensity? A. Who Do We Study? The empirical setting is a panel of publicly listed U.S. financial firms observed in filing and accounting data. The broader design targets a sample of roughly 100 firms, with a possible extension to 300 firms for robustness or appendix analysis. In the current draft, the AI-specific filing analysis uses a focused sample of five large institutions: Bank of America, BNY Mellon, JPMorgan Chase, S&P Global, and State Street. Firm-years are assigned to the three broad technology eras using calendar-year cutoffs: Computerization ( 1985-2000 ), Index investing ( 2000-2015 ), and AI / automation ( 2015-present ). These cutoffs are stylized, but they provide a transparent way to compare broad phases of technological change in finance. B. What Is the Productivity Measure? The main productivity outcome is revenue per employee, which captures broad operating productivity and is available across a wide set of firms. Secondary outcomes include labor expense per employee and, where relevant, assets under management per employee. The AUM-based measure is especially informative for asset managers, custodians, and related firms in which scale can expand without a proportional increase in headcount. Together, these outcomes distinguish different channels through which automation may matter. Revenue per employee captures overall operating productivity. AUM per employee captures scale in information-processing and oversight-intensive businesses. Labor expense per employee provides a direct, though incomplete, proxy for whether organizational automation is translating into near-term labor-cost compression. C. How Is Task Automation Intensity Measured Across Eras? The era-specific automation measures are built from annual reports and 10-K filings. For the computerization era, the relevant language includes terms related to computers, terminals, spreadsheets, electronic trading, digitization, workflow systems, and back-office automation. For the