Only 8% of Fund Managers See Higher Returns From AI, Report Finds

Despite accelerating integration of artificial intelligence across international asset management firms, only eight percent of institutional fund managers report measurable gains in portfolio investment returns, according to joint research by Morgan Stanley and Oliver Wyman. The empirical survey results highlight a substantial structural gap between the widespread adoption of automated technology and its tangible capacity to generate superior financial returns for clients.

Industry Adoption Breadth and the Alpha Deficit

The findings were published in a September 2026 BluePaper report titled Breaking from the Pack: The Race for Above-Market Growth, which evaluated technological disruption across financial asset managers. Drawing on comprehensive data from the Mercer 2026 AI in Asset Management Survey, the study noted that 55 percent of asset managers have embedded machine learning into at least one stage of their operations. Investment returns have not yet realized widespread structural gains from automation, with just eight percent of surveyed firms documenting verifiable alpha benefits, the report stated. Industry analysts emphasized that while advanced analytics accelerate document ingestion, commercial evidence connecting machine learning to consistent market outperformance remains scarce, according to The Economic Times.

Operational Efficiency Versus Corporate Cost Reductions

Survey findings revealed that 69 percent of participating institutions achieved demonstrable enhancements in internal operational workflows, while 55 percent cited sharper research insights. Cost reductions failed to materialize at a comparable scale, with only 24 percent of participating firms reporting lower overall operational expenditure. The disparity suggests that while machine learning tools accelerate internal administrative cycles, significant licensing fees and computational infrastructure costs offset baseline labor savings across mid-sized asset management firms.

Workflow Concentration in Research Versus Portfolio Construction

Technological adoption proved heavily concentrated within preliminary equity research, where 49 percent of surveyed investment organizations currently integrate machine learning models, followed by data processing at 40 percent. Asset managers remained notably hesitant to delegate fiduciary portfolio construction, with only 13 percent deploying algorithms for asset weighting and just seven percent automating trade execution. Whether expanded research capacity and automated synthesis ultimately translate into persistent alpha across market cycles remains unproven, the report said. Investment desks in the corporate sector continue to confine predictive algorithms to screening unstructured data rather than granting autonomous execution mandates, according to reporting by The Economic Times.

Productivity Paradox in Enterprise Financial Strategy

The joint research paper cautioned that faster data processing and enhanced analyst output do not automatically lead to superior fund performance across institutional mandates. Productivity gains do not automatically translate into enterprise-wide financial impact, it stated, underscoring that raw informational speed cannot substitute for differentiated fundamental judgment. The report concluded that institutional investment firms must fundamentally restructure organizational decision-making workflows rather than merely purchasing commercial language model subscriptions.

Operational gains are different from investment alpha

The report's eight-percent figure is a share of surveyed managers reporting improved returns, not evidence that AI loses money for the remaining 92 percent or that every firm measured its performance in the same way. Portfolio returns are affected by market direction, benchmarks, fees and risk; attributing a return gain specifically to a new AI tool requires a rigorous comparison with a reasonable counterfactual. Faster research and document processing are easier to observe than incremental investment performance.

For asset-management clients, the relevant questions are whether AI-generated signals survive out-of-sample testing, whether human oversight catches hallucinations or biased data, and how compliance rules protect confidential investor information. Improved workflows can still be valuable even if they do not immediately translate into higher alpha. Conversely, a favourable survey response does not validate any particular trading strategy. The report is a snapshot of respondent experiences, not a promise of future investment returns.