AI is reshaping equity research — from financial modeling to filing analysis to idea generation. Here's what's actually changing, what isn't, and what the data shows.
Equity research as an industry has been under pressure for over a decade. MiFID II regulations in Europe (2018) unbundled research from trading commissions, forcing buy-side firms to pay explicitly for research and sell-side firms to justify the cost of producing it. The number of sell-side analysts covering public companies has declined — S&P Global Market Intelligence reported a 10% reduction in sell-side analyst headcount between 2017 and 2022.
At the same time, the volume of data available to analysts has exploded. SEC EDGAR alone contains over 21 million filings. Companies are filing longer 10-Ks — the average length of an annual report for S&P 500 companies increased from approximately 42,000 words in 2000 to over 60,000 words by 2023, according to research by Loughran and McDonald. Earnings calls, investor presentations, ESG disclosures, and alternative data sources have all expanded.
Fewer analysts processing more information. That's the environment AI entered.
The applications of AI in equity research fall into several categories, each at a different stage of maturity:
This is the most mature application. AI tools can now build three-statement financial models — income statement, balance sheet, and cash flow statement — directly from SEC XBRL data. The historical data is extracted automatically, the statements are linked, and the model is ready for the analyst to add projection assumptions.
The data supports the efficiency claim. Building a three-statement model manually takes 8-40 hours depending on experience level and model complexity. Automated construction reduces the data entry phase to near-zero. The analyst's time shifts entirely to assumption-setting and analysis.
10-K filings for large companies routinely exceed 100 pages. AI can process these documents and surface key changes: new risk factors, shifts in revenue recognition, changes in segment reporting, updated guidance language. This doesn't replace reading the filing — it helps analysts prioritize where to focus.
A 2024 study by Cao, Jiang, Yang, and Zhang published in the Journal of Financial Economics found that AI-generated summaries of earnings calls could predict subsequent stock returns, suggesting that AI can extract analytically relevant information from unstructured financial text.
AI tools can screen the full universe of SEC-filing companies — approximately 7,000 — against specific financial criteria in ways that go beyond traditional quantitative screening. Instead of filtering on P/E ratios or revenue growth, an analyst can query for companies where management discussion mentions "pricing power" alongside margin expansion, or where capex as a percentage of revenue has been declining for three consecutive years.
This is particularly valuable for buy-side analysts who need to find new ideas outside their existing coverage universe. The entire SEC filing database becomes searchable, not just the companies they already follow.
Building a comps table traditionally requires pulling financial data for 10-20 peer companies, standardizing the metrics, and calculating multiples. AI can automate the data extraction and standardization, producing a comps table from SEC data for any set of public companies in minutes rather than hours.
Several data points illustrate how AI adoption is progressing in equity research:
Bloomberg's 2024 survey of institutional investors found that 56% of buy-side firms reported using AI tools for some aspect of their investment process, up from 33% in 2022.
Deloitte's 2024 investment management outlook reported that 75% of asset managers surveyed were investing in AI capabilities, with financial analysis and research cited as the top use case.
The CFA Institute's 2023 Future of Finance report found that 54% of investment professionals expected AI to significantly change their workflow within three years, with data processing and financial modeling cited as the areas most likely to be affected.
Academic research has begun quantifying AI's analytical capabilities. A 2023 study by Kim, Muhn, and Nikolaev at the University of Chicago found that GPT-4 could analyze financial statements and predict future earnings changes with accuracy comparable to professional analysts — and outperformed them when analysts' predictions appeared to be biased.
Despite the hype, several core aspects of equity research remain fundamentally human:
Investment judgment. Deciding whether a stock is a buy at the current price requires synthesizing financial data, industry context, management quality assessment, competitive dynamics, and macro conditions into a probabilistic view about the future. AI can organize and present information, but the conviction call is human.
Relationship-based information. Much of what makes top analysts effective is access to management teams, industry contacts, and channel checks. These relationships produce insights that aren't in any filing or dataset.
Creative thesis development. The most valuable investment ideas often come from connecting non-obvious dots — recognizing that a regulatory change in one industry will benefit a company in another, or that a technology shift will create winners and losers in unexpected ways. This kind of lateral thinking is difficult to automate.
Communication and persuasion. Sell-side research is ultimately a communication product. Buy-side analysis needs to be presented to portfolio managers and investment committees in a way that builds conviction. Writing compelling investment narratives remains a human skill.
The pattern is consistent with previous technology shifts in finance: the job doesn't disappear, but the task composition changes.
Before AI: An equity research analyst's week might break down as 30% data gathering, 25% model building and maintenance, 20% reading filings and research, 15% writing reports, and 10% talking to management and industry contacts.
With AI tools: The data gathering and model maintenance shrinks dramatically. The reading, analysis, and communication expand to fill the time. The analyst covers more companies, spends more time on each investment thesis, or both.
This shift mirrors what happened when Bloomberg terminals became widespread in the 1990s. Analysts didn't spend less time working — they spent the same time doing higher-value work. The analysts who adopted the tools early gained a coverage and speed advantage over those who resisted.
For sell-side equity research, AI has particular implications. The MiFID II-driven economic pressure means research departments need to produce more coverage with fewer analysts. AI tools that automate model construction and data extraction allow a single analyst to maintain coverage of more companies.
The value proposition of sell-side research is also shifting. When any buy-side analyst can generate a three-statement model from SEC data in minutes, the sell-side model is no longer the product — it's the insights, industry expertise, and management access that the model supports. Firms that recognize this are investing in differentiated content rather than competing on data delivery.
For buy-side firms, AI's primary value is in expanding the investable universe. When model construction and maintenance are no longer the bottleneck, analysts can evaluate more opportunities and maintain views on more companies.
This is particularly valuable for:
The trajectory of AI in equity research points toward deeper integration rather than replacement. The tools are becoming better at:
Each of these capabilities shifts analyst time further toward judgment and away from processing. The analysts who adapt — learning to use AI tools as leverage rather than viewing them as threats — will be the ones who produce the best research and generate the most alpha.
The firms that don't adapt will find themselves at an increasing disadvantage: slower to update models, narrower in coverage, and spending more of their expensive human capital on work that machines do better.