A Stanford research team led by Ed deHaan built an AI stock analyst and backtested it over 30 years (1990-2020). It outperformed 93% of human mutual fund managers, generating roughly 600% excess alpha.

Then 2025 happened.

Man Group’s AHL — one of the world’s largest systematic hedge funds — dropped 15% by April 2025 as “garbage rallies” in heavily-shorted stocks wrecked momentum algorithms. Pierre Andurand’s flagship commodity fund fell over 50%. Goldman Sachs prime services estimated quant equity managers lost 4.2% in just June-July alone.

Man Group recovered to finish the year up 5%. Many pure-AI strategies didn’t.

A separate Harvard-led study (Cohen, Lu, Nguyen — NBER, February 2026) found that AI can predict 71% of mutual fund trading decisions. But it’s the other 29% — the unpredictable trades — where actual alpha lives.

The gap between a 30-year backtest and a single bad quarter is the story of AI in capital markets. The algorithm works until it doesn’t. And “until it doesn’t” is where human judgment earns its premium.

I analyzed 20 capital markets roles across research, asset management, quantitative trading, and executive leadership. The results reveal an industry where AI has penetrated deeper than almost anywhere else — and yet the most human roles are also the most valuable.

The Numbers

Category # of roles Avg replacement rate
Heavy AI assistance (50-70%) 7 57%
Limited AI assistance (25-50%) 8 36%
Irreplaceable (<25%) 4 17%
Net job growth 1 -20% (growing)

Industry-wide AI replacement rate: approximately 38% — lower than banking (47%), marketing (53%), or accounting (48%). Why? Because capital markets roles disproportionately require judgment under uncertainty, relationship networks, and the ability to be wrong in ways that matter.

But that average hides the most extreme dumbbell effect I’ve seen in any industry.

The Wall Street Dumbbell

In every industry I’ve analyzed, AI hits juniors harder than seniors. In capital markets, the split is brutal.

Junior analysts (1-3 years experience): AI automates 80-90% of their data collection, model building, and report drafting. Goldman Sachs reported 40-70% reduction in junior data collection tasks after deploying its AI research tools. Initial hiring is already falling — entry-level finance and accounting job postings dropped over 50% since 2022 (Adzuna).

Senior professionals (8+ years): AI makes them more productive, not less necessary. Fidelity Investments deployed AlphaSense and reduced analyst headcount by 15% — but each remaining analyst now covers 5 industries instead of 3. Individual output rose. Individual value rose.

The middle (4-7 years): squeezed hardest. Too senior to be pure data processors, too junior to have built the relationship networks and judgment that justify their cost. This cohort faces the deepest career risk in all of finance.

Bloomberg reported in January 2025 that Wall Street analyst pay has dropped 30% — inflation-adjusted — over the past decade. Sell-side research headcount at the top 15 banks: approximately 4,600 a decade ago, roughly 3,000 today. That’s a 35% cut, and AI is accelerating it.

REPLACE Tier: The Roles Algorithms Are Absorbing

Buy-Side Investment Analysts — 55-65% automation

The highest-risk research role in capital markets. AlphaSense ($4B+ valuation, covering 40% of the largest asset managers) and FactSet AI automate financial statement parsing, competitive analysis, earnings call summaries, and sentiment analysis. JPMorgan’s AI research platform now serves 5,000 analysts.

Stanford’s AI professor explicitly warned: “Junior buy-side analyst positions are in jeopardy.” Goldman reported AI-generated report drafts are 40% faster, though still scoring 25% lower on “insight depth” compared to top human analysts.

Fidelity’s experience is a preview: 15% fewer analysts, each covering nearly double the industries. Expect 30-40% reduction in junior buy-side analyst headcount by 2028.

Algorithmic Trading Engineers — 60% automation

The execution layer is fully AI-driven. Algorithms now handle 70-80% of US equity trading volume daily. The global algorithmic trading market: $21 billion in 2024, projected $43 billion by 2030 (CAGR 12.9%).

But strategy design, regulatory compliance, and crisis judgment remain human. Jane Street learned this the hard way: India’s SEBI froze $566 million of the firm’s assets in July 2025 over allegations of Bank Nifty manipulation. The algorithms executed perfectly. The compliance question was entirely human.

Compliance Directors — 55-70% automation

Securities compliance is more automatable than banking compliance because it’s heavily rule-based: trade surveillance, position limit monitoring, disclosure filing. RegTech tools automate the mechanical work. But EU AI Act enforcement (August 2, 2026) and new FINRA algorithmic trading rules add layers of AI-specific oversight that require human interpretation.

Asset Allocation — Retail Layer: 85-95% automated

Robo-advisors have effectively replaced human retail asset allocation. Vanguard Digital Advisor: $311 billion. Wealthfront: $75 billion. Schwab Intelligent Portfolios: $81 billion. Betterment: $46 billion. Total robo AUM has crossed $1 trillion globally. Service fees dropped from 1% (human) to 0.25% (robo).

The institutional layer is different — macro judgment, geopolitical risk, and cross-asset framework design still require humans. But the retail advisory role is, for practical purposes, automated.

AMPLIFY Tier: Fewer People, Bigger Decisions

Quantitative Analysts — 55% automation

Goldman’s cash equity desk went from 600 traders in 2000 to 2 traders plus 200 engineers. Two Sigma (now ~$70 billion AUM) reports that 80% of data processing tasks are AI-completed. Quant roles are transforming, not vanishing.

But the “quant winter” of 2025 proved the limits. Multiple pure-AI strategies suffered 30%+ drawdowns when market microstructure shifted. Two Sigma’s own co-CEO resigned in March 2026 amid governance challenges — a reminder that the human layer around AI matters as much as the AI itself. As AXA Investment Managers put it in their 2026 outlook: “Humans remain firmly in control. Domain knowledge and overfitting prevention are the real alpha, not the AI models themselves.”

The top quant analysts — those creating novel alpha hypotheses — command 15-20% higher salaries. Entry-level quant positions are shrinking 20-30%.

Portfolio Managers — 40-55% automation

BlackRock’s Aladdin platform now monitors approximately $25 trillion in assets across 5,000+ institutional clients. AI handles real-time risk monitoring, portfolio optimization, and trade execution. A Harvard Business School study found AI can predict 71% of fund managers’ “predictable” trading decisions.

But the other 29% is where fund managers earn their fees. In 2025’s volatile markets, human-managed macro funds outperformed pure-AI strategies. Bridgewater’s AIA Labs — its dedicated AI fund — surpassed $5 billion AUM and returned +11% in 2025 by combining AI signals with human macro judgment.

The contrast with Two Sigma’s governance crisis is instructive: AI succeeds when the organization supports it. It fails when humans fight over control.

ESG Investment Analysts — 50-60% automation

RepRisk processes 500,000+ documents daily across 22 languages. MSCI ESG Manager covers 8,500+ companies. Sustainalytics updates ratings weekly instead of quarterly. The data collection layer is heavily automated.

But EU SFDR revision (2026) and the new EU ESG ratings regulation (effective July 2, 2026) create fresh demand for human interpretation. “Greenwashing” identification — now at risk of becoming “AI-washing” — requires qualitative judgment. BlackRock’s ESG team expanded after deploying AI tools, redirecting saved time into “active ownership” (shareholder engagement and voting). ESG regulation is growing faster than AI can simplify it.

Equity Research Analysts — 30% automation

Equity research is dying — but not because of AI. MiFID II (2018) separated research fees from trading commissions, gutting sell-side revenue. AI is accelerating a structural decline already underway. AlphaSense compresses industry scan time from 6-8 hours per week to minutes. Morgan Stanley’s AskResearchGPT covers 10,000+ reports.

The survivors are specialists with differentiated insights, expert networks, and client trust — none of which AI replicates. Buy-side firms are building internal research capabilities (cheaper with AI), reducing sell-side dependence. Independent sell-side research firms will decline another 40% by 2030.

IRREPLACEABLE Tier: Where Human Judgment IS the Product

Macro Economists — 12% automation

The single most AI-proof research role in capital markets. MIT’s Acemoglu calculated that AI’s 10-year cumulative impact on total factor productivity is just 0.53-0.66%. The ECB’s Project Spectrum reduced short-term inflation forecast errors by 18% — but couldn’t predict policy turning points.

Bridgewater’s Ray Dalio explicitly stated that macro research is the last domain where AI “assists but does not lead.” Building a narrative about why the current economic cycle differs from history — integrating economics, politics, demography, and geopolitics — is irreducibly human.

C-Suite (CEO, CIO) — 10-15% automation

The CEO of a securities firm and the Chief Investment Officer both sit at the intersection of strategy, regulation, and trust. Fortune’s December 2025 investigation concluded: “AI really killing finance jobs may be more hype than takeover.” The C-suite directs AI transformation. It doesn’t get replaced by it.

HFT System Engineers — 35% task replacement, 5-10% job elimination

The paradox role. Execution has been 100% automated for two decades — nanosecond-level, FPGA-driven, no human in the loop. But the system architecture that enables it (hardware selection, co-location optimization, network topology, resilience design) is one of the most specialized engineering disciplines in existence. Globally, fewer than 5,000 people have top-tier HFT system design capability. Salaries reflect it: $200K-$500K+. AI helps them test and monitor. It cannot replace them.

The One Growing Role

Out of 20 capital markets roles I analyzed, exactly one has a negative replacement rate — meaning AI is creating more jobs, not fewer.

Alternative Data Analysts — Net growth (-20% replacement)

The alternative data market was valued at $11.65 billion in 2024. It’s projected to reach $135.7 billion by 2030 — a 52-63% CAGR, depending on the forecast (Grand View Research). Over 60% of top hedge funds (AUM $1B+) already purchase alternative data services.

Point72’s alternative data team grew from 5 people in 2019 to 35 in 2024 — the fastest-growing research sub-team in the firm. BlackRock acquired Preqin for $3.2 billion in March 2025 to integrate alternative fund data into Aladdin.

Why is this role growing? Because AI dramatically lowered the cost of processing non-traditional data (satellite imagery, credit card transactions, app usage, job postings). That made more types of data investable — which created demand for more analysts who can translate raw signals into trading strategies. AI is the engine of this role’s growth, not its threat.

The skill: discovering data sources nobody else has found, and designing creative strategies around them. AI processes the data. Humans decide what’s worth processing.

The Governance Lesson

The most instructive contrast in 2026 capital markets:

Bridgewater’s AIA Labs surpassed $5 billion AUM and returned +11% in 2025. It combines AI signals with deep macro judgment, overseen by co-CIO Greg Jensen. The AI works because the humans around it are aligned and capable.

Two Sigma — managing ~$70 billion — saw its co-CEO Scott Hoffman resign on March 31, 2026, citing “governance challenges.” The firm’s AI technology is world-class. Its human governance is fracturing.

The lesson: the differentiator in AI-era capital markets isn’t the model. It’s the organization around the model. AI amplifies whatever the humans bring — competence or dysfunction.

What This Means For You

If you’re in capital markets:

  1. The dumbbell is real and accelerating. Junior roles are being compressed by 30-50%. Senior roles are becoming more leveraged. If you’re in the 4-7 year middle, the clock is ticking to build either deep specialization or relationship networks.

  2. Backtests are marketing. Live performance is truth. Stanford’s AI beat 93% of fund managers on paper. Man Group lost 15% in a single quarter in live markets. Anyone selling you “AI-powered alpha” without live track records is selling a backtest.

  3. The only growing role is the one that didn’t exist a decade ago. Alternative data analysis is the sole net-growing job in capital markets. The pattern: AI doesn’t just destroy old roles. It creates new ones — and the new ones often require more human creativity, not less.

  4. Governance beats technology. Bridgewater’s AIA Labs ($5B, +11%) thrives while Two Sigma’s co-CEO resigns. The firms that win the AI era aren’t the ones with the best models. They’re the ones with the best humans around the models.

Capital markets aren’t being replaced by AI. Capital markets are being reorganized around AI — and the reorganization favors judgment, relationships, and the courage to disagree with the algorithm.

This is part of my 119-industry AI replacement analysis series, based on the Replace / Amplify / Emerge framework. I’ve analyzed 20 capital markets roles across research, asset management, quantitative trading, and executive leadership.

Previously: HR, Software/Tech, Finance (overview), FinTech, Sales & Marketing, Accounting/Audit/Tax, Banking & Financial Services.

Follow for the next analysis: Legal.

Sources

  • Stanford GSB — Ed deHaan et al., AI Stock Analyst Outperforms 93% of Fund Managers (30-Year Backtest, 2025): https://news.stanford.edu/stories/2025/06/ai-stock-analyst-analysis-performance-human-mutual-fund-managers
  • Harvard / NBER — Cohen, Lu, Nguyen: AI Predicts 71% of Fund Trades (Feb 2026): https://www.bloomberg.com/news/articles/2026-02-24/harvard-led-study-says-ai-can-predict-71-of-active-fund-trades
  • Bloomberg — Wall Street Analyst Pay Drops 30% (Jan 2025): https://www.bloomberg.com/news/features/2025-01-08/wall-street-analyst-pay-drops-30-as-banks-slash-equity-research
  • Man Group — AHL -15% YTD April 2025 (Quant Winter): https://www.bloomberg.com/news/articles/2025-04-11/man-group-hedge-funds-losing-up-to-15-this-year-show-quant-pain
  • MSCI — Summer 2025 Quant Fund Wobble Analysis: https://www.msci.com/research-and-insights/blog-post/unraveling-summer-2025s-quant-fund-wobble
  • Bridgewater — AIA Labs Surpasses $5B AUM, +11% in 2025: https://www.hedgeco.net/news/03/2026/bridgewater-dalios-principles-to-algorithmic-intelligence-the-road-to-5billion.html
  • Two Sigma — Co-CEO Scott Hoffman Resigns March 2026: https://www.hedgeweek.com/two-sigma-co-ceo-resigned-due-to-governance-challenges/
  • Two Sigma — AI in Investment Management 2026 Outlook: https://www.twosigma.com/articles/ai-in-investment-management-2026-outlook-part-i/
  • BlackRock — Aladdin Monitors ~$25T in Assets: https://en.wikipedia.org/wiki/Aladdin_(BlackRock)
  • BlackRock — Preqin $3.2B Acquisition (March 2025): https://www.blackrock.com/corporate/newsroom/press-releases/article/corporate-one/press-releases/blackrock-completes-preqin-acquisition
  • AlphaSense — $4B+ Valuation, 40% of Top Asset Managers: https://www.prnewswire.com/news-releases/alphasense-to-join-forces-with-tegus-increases-latest-valuation-to-4b-302169502.html
  • Goldman Sachs — 600 Traders to 2 + 200 Engineers: https://www.technologyreview.com/2017/02/07/154141/as-goldman-embraces-automation-even-the-masters-of-the-universe-are-threatened/
  • Jane Street — SEBI Freezes $566M (July 2025): https://www.cnbc.com/2025/07/04/indian-regulator-bars-us-trading-firm-jane-street-from-accessing-securities-market.html
  • Grand View Research — Alternative Data Market $11.65B→$135.7B: https://www.grandviewresearch.com/press-release/global-alternative-data-market
  • Grand View Research — Algorithmic Trading Market $21B→$43B (CAGR 12.9%): https://www.grandviewresearch.com/press-release/global-algorithmic-trading-market
  • EY — 95% WAM Firms Adopted GenAI, 97% Minimal Headcount Change: https://www.ey.com/en_us/insights/wealth-asset-management/gen-ai-in-wealth-asset-management-survey
  • MIT / Acemoglu — AI TFP Impact 0.53-0.66% Over 10 Years
  • ECB Project Spectrum — GenAI Inflation Forecast Error -18%
  • AXA IM — “Humans Remain Firmly in Control” (Quant Outlook 2026): https://core.axa-im.com/investment-strategies/equities/insights/what-quant-investing-looks-2026-data-ai-and-human-judgment
  • Fortune — “AI Killing Finance Jobs: More Hype Than Takeover” (Dec 2025): https://fortune.com/2025/12/21/is-ai-killing-finance-and-banking-jobs-experts-say-wall-street-layoffs-hype-than-takeover/
  • Adzuna — Entry-Level Finance Jobs -50.8% Since ChatGPT Launch: https://www.techerati.com/news-hub/entry-level-jobs-fall-by-32-since-emergence-of-chatgpt/
  • Robo-Advisor AUM — Vanguard $311B, Wealthfront $75B, Schwab $81B, Betterment $46B: https://investingintheweb.com/brokers/the-largest-robo-advisors-by-aum/
  • Renaissance Technologies — Medallion Fund 66% Annualized Pre-Fees: https://en.wikipedia.org/wiki/Renaissance_Technologies
  • SEC — Pattern Day Trader Rule Eliminated (April 14, 2026): https://coinalertnews.com/news/2026/04/15/sec-eliminates-pattern-day-trader-rule
  • EU — ESG Ratings Regulation Effective July 2, 2026
  • EU AI Act — High-Risk System Enforcement August 2, 2026: https://ai-act-service-desk.ec.europa.eu/en/ai-act/timeline/timeline-implementation-eu-ai-act