Age of AI II: Hedge Funds Struggle to Make AI Routine

The Age of AI II report, published June 18, 2026, draws on a global online survey of more than 100 hedge fund managers conducted in the first half of 2026 and follow-up interviews with executives and heads of research. The report finds widespread purchase of AI tools during the prior year but a gap between ownership and day-to-day application. Many respondents indicated models remain in research labs or on analysts’ desktops rather than embedded in trading systems, portfolio construction or risk processes.

Respondents identified several recurring barriers. Technical integration was cited as a problem when new models must link with legacy execution systems and data platforms. Incomplete or poorly organised historical data limited the ability to train and validate models. Governance and compliance requirements led risk teams to demand explainability, audit trails and stricter change controls before models could be moved to live trading. Firms also reported difficulty hiring staff with both trading domain knowledge and machine learning skills and challenges demonstrating a clear return on investment beyond small experiments.

Adoption patterns varied by strategy and firm size. Systematic and quantitative managers and larger firms were more likely to report regular use of AI in research and execution. Many discretionary and smaller managers reported a slower transition from experimentation to production. Firms with dedicated in-house data and engineering teams reported faster and broader rollouts than those relying mainly on external vendors.

Managers that regularly use AI described narrower, measurable use cases. Examples include signal generation, trade idea screening, risk monitoring and operational automation. These firms invested in data infrastructure to store, clean and version inputs; built engineering pipelines to move models from prototype to production; and formed cross-functional teams combining quantitative researchers, data engineers and risk staff. Governance steps included model documentation, performance monitoring and staged deployment to limit operational risk.

The report highlights measurement and incentives. Firms tracking production metrics such as live-model performance, latency and error rates were able to iterate faster and reallocate resources to projects that met production targets. Several managers adjusted compensation and promotion criteria to reward staff who delivered production outcomes rather than only academic papers or improved backtests.

Interviews described trade-offs between third-party and in-house approaches. Some firms adopted vendor models for speed and cost control and added proprietary preprocessing, post-processing and governance layers to meet compliance needs. Other firms reported that building internal capabilities required multi-year investment in staff and systems and involved higher fixed costs, which limited that option to larger managers or those making explicit strategic bets.

Risk officers across multiple firms declined to approve black-box models for certain portfolio decisions without stronger diagnostics and controls. Respondents mentioned regulatory attention to model risk and data handling as a factor shaping governance requirements.

The report, reflecting mid-2026 data, records that many hedge funds now have AI capability but that integration into regular trading and risk workflows remains uneven. The report documents the practices used by firms that have moved models into production.