Howard Marks: AI Will Eliminate Most Fund Managers, But the Greatest Investors Will Thrive

Artificial intelligence will likely push most fund managers out of asset management, according to Howard Marks, but the survivors will be those who excel in the areas where AI is weakest: judgment, intuition, and navigating non-linear situations without historical precedent.

In Bloomberg reporting published today, the Oaktree Capital Management co-chairman laid out a paradox that challenges the prevailing narrative about AI’s dominance in finance: the technology’s greatest strength—absorbing vast amounts of data and identifying patterns—is also its fundamental limitation in investing, where the future rarely mirrors the past.

According to Marks, who wrote his latest note with assistance from Anthropic’s Claude AI model, the money managers who remain in the industry will be expert in assessing management skill, evaluating the importance of new products, and handling other qualitative factors where AI lags. They will also excel at navigating situations that don’t easily match prior patterns—precisely the scenarios that define exceptional investing.

The Pattern Recognition Trap

The distinction Marks draws hinges on understanding what AI does exceptionally well versus what investing actually requires. Machine learning algorithms excel at identifying patterns in historical data—the same capability that powered Renaissance Technologies’ legendary Medallion Fund to average 66% annual returns before fees from 1988 to 2018, according to Wikipedia.

But Renaissance’s success proves Marks’ point rather than contradicts it. The fund’s co-CEO Robert Mercer revealed the fund was right only about 50.75% of the time—barely better than a coin flip. What mattered was executing millions of trades annually, exploiting tiny inefficiencies at massive scale. Research indicates the firm earned just 0.01% to 0.05% per trade, meaningless for a single transaction but extraordinary across millions.

This approach works brilliantly for quantitative arbitrage but fails catastrophically when historical patterns break down. As Michigan Journal of Economics notes, AI cannot predict unexpected events without historical patterns—a CEO resigning due to family emergencies, sudden regulatory shifts, or black swan events that define market turning points.

The Index Fund Parallel

Marks’ argument echoes a transformation already underway. Just as index funds displaced countless active managers who couldn’t justify their fees, AI will eliminate another layer of middling performers. Industry research shows passive products attracted 70% of total global mutual fund and ETF net flows in 2023—about $920 billion—while fee compression continues to accelerate.

But the parallel extends further. Warren Buffett survived the index fund revolution not by competing on processing speed or data coverage, but by making non-consensus judgments about business quality, management integrity, and competitive moats. According to Hedge Fund Alpha, Marks argued in a previous interview that investing “is an art form” where “some people have insight” and “can make these qualitative, subjective distinctions” that most cannot.

The U.S. Bureau of Labor Statistics reports approximately 868,600 financial managers held jobs in 2024. If Marks is correct, a significant portion of these roles—particularly those focused on data synthesis and pattern matching—face displacement similar to what clerical workers experienced during computerization.

Where Human Judgment Still Matters

The investment scenarios where humans retain advantage share common characteristics: they involve assessing factors with limited historical data, understanding second-order effects, and making calls that seem irrational until they’re vindicated.

Consider Marks’ own investment philosophy, detailed in his decades of client memos. In his recent note “Is It a Bubble?”, he wrote that identifying market excess is “just a matter of judgment.” No algorithm trained on historical data could have called the 2000 tech bubble or 2007 credit crisis at the right moment—both required recognizing that “this time” was not, in fact, different, despite novel circumstances.

Investment management research confirms that AI cannot fully replace human intuition, experience, and ethical judgment. While AI processes large datasets and detects patterns with exceptional accuracy, it struggles with the “black box” problem—the difficulty in interpreting and explaining decisions made by complex models.

The qualitative factors Marks identifies—management assessment, product importance, strategic positioning—require understanding human behavior, organizational dynamics, and competitive strategy in ways that resist quantification. An AI can tell you a CEO’s compensation structure and tenure; only human judgment can assess whether they have the fortitude to make unpopular decisions during a crisis.

The Uncomfortable Middle Ground

The asset management industry now faces bifurcation. At one end, pure quant strategies deploying AI at scale; at the other, elite fundamental investors making non-consensus calls. The uncomfortable middle—traditional active managers who blend some quantitative screening with conventional analysis—faces extinction.

This group historically justified 1-2% management fees by providing “sophisticated” analysis that was really just faster data processing. As HedgeCo Insights reports, AI now performs these tasks at roughly half the cost of a junior human analyst.

Amundi Research Center notes that while AI improves investment research by cutting data noise and analyzing unstructured data, “historical patterns may not reliably predict future market behaviour, algorithms can overlook critical factors, and machines can find spurious relationships that are just aberrations.”

What to Watch

The next 24 months will clarify whether Marks’ thesis holds. Watch for three signals: First, fee structures at traditional active managers—if they start offering lower-fee, AI-augmented products, it signals capitulation. Second, recruiting patterns at elite funds—are they hiring fewer MBAs and more specialists in unique domains (healthcare, energy transition, geopolitics)? Third, performance dispersion among hedge funds—if the gap between top and median performers widens, it confirms that skill has become more, not less, important.

Renaissance Technologies’ other funds—RIEF and RIDA—provide instructive precedent. While Medallion achieved legendary returns exploiting micro-inefficiencies, these larger funds targeting institutional investors generated solid but unremarkable performance. Analysis shows RIEF was up 22.5% through October 2024 while RIDA gained 15.6%—excellent but not extraordinary. The scalable strategies produced good returns; the greatest excess returns remained locked in the employee-only Medallion Fund where human insight still guided model development.

Marks himself demonstrates the path forward. His latest memo was written with AI assistance, but the judgment about what matters—the distinction between mean-reversion and inflection bubbles, the assessment of when optimism crosses into mania—remains distinctly human. The investors who thrive won’t be those who reject AI, but those who deploy it while reserving the crucial decisions for capabilities machines cannot yet replicate: wisdom born from experience, courage to act against consensus, and the judgment to know when the future will not resemble the past.