2026 Benchmarking Survey Report
This survey was conducted in partnership with Addepar.
We asked 54 asset owners, who together oversee more than $3 trillion in assets, how they use AI today, what holds adoption back, and where they expect it to matter next. Explore the findings below.
Executive summary
80% of respondents use AI tools every day, and 98% reach for general-purpose assistants such as ChatGPT, Claude, and Gemini. Use concentrates in operational workflows (80%) and manager due diligence (69%). Scaling usage across the organization is common, yet only 7% have fully embedded AI into core investment or operational processes.
On governance, 39% are developing a formal policy and 35% have one fully in place. Data confidentiality leads the barrier list at 69%, and primary concern divides between output reliability (44%) and data confidentiality (43%). 94% expect their AI usage to rise over the next 12 to 24 months, and no respondent expects it to fall. Asked what would accelerate adoption, 80% point to better workflow integration.
Respondent profile
Of the 54 respondents, endowments (43%), family offices (22%), and foundations (19%) make up 84%, with OCIOs, pensions, and other vehicles forming the balance. Investment operations, technology, and reporting roles account for 54%, and research and due diligence for 19%. 76% manage between $1bn and $20bn, and 93% are based in the Americas.
AI use today
80% of respondents use AI daily, and 94% use it at least weekly. General-purpose assistants dominate the toolkit at 98%, well ahead of platform-embedded features, coding tools, and internally built copilots. Use concentrates in operational workflows, manager due diligence, portfolio monitoring, and investment committee materials.
Maturity tells a sharper story than frequency alone. Expanding AI across the organization has become common practice, while embedding it in the core investment and operational processes that run the firm remains the exception, reached by fewer than one in ten.
Barriers, concerns, and governance
Data confidentiality and security leads the barrier list at 69%, followed by integration with existing systems, then internal expertise and data quality. Asked for a single primary concern, respondents split almost evenly between the reliability of AI output and data confidentiality, which together account for 87% of first responses. Inside teams the posture is constructive, with enthusiasm and rising adoption the common reports. Formal governance is largely in motion, set at some firms and in development at more.
Expected impact and outlook
Operational efficiency and automation leads expected impact at 69%, with portfolio monitoring and data management close behind. The outlook points one way: 94% expect their use of AI to rise over the next 12 to 24 months, and no respondent expects it to fall. Better integration with existing workflows is the clear top accelerant at 80%, ahead of internal training and security controls.
In their words
Thirty-six respondents described where AI has delivered measurable value. Their answers cluster in drafting, document summarization, data work, and due diligence, the same operational core the closed questions rank highest.
Fifteen respondents addressed what the industry conversation underplays. Data foundations and connectivity came up most, alongside single mentions of business-led adoption and executive sponsorship, ROI discipline, and preparing teams for AI-enabled workflows.
What this means
Adoption is settled, so the real question is depth. Broad daily use has not yet reached the core systems where investment and operations run, and closing that gap is the work ahead. The place to start is where activity already concentrates, in operations, diligence, and reporting.
Security and governance set the terms. Confidentiality and the reliability of output are the twin checks on how fast teams can move, and formal policy is being written alongside adoption rather than ahead of it.
Integration decides how far this goes. It is the constraint that appears on both sides of the ledger, slowing adoption now and standing as the single change most say would speed it, and respondents name the same limit themselves, that disconnected systems leave AI with a partial view. The efficiency most expect arrives when AI connects to the systems where the data already sits.
This report presents the results of a survey of 54 institutional asset owners, fielded in July and August 2026. Respondents span endowments, foundations, family offices, OCIOs, pensions, and related vehicles. The sample concentrates in the Americas and in the $1bn to $20bn asset range, and its composition is described in the respondent profile.
Unless a chart notes otherwise, the base for each question is the full sample of 54, and every respondent completed each single-select question, so item non-response does not affect those results. Single-select questions sum to 100% within rounding. Multiple-response questions, identified where they appear, invited respondents to select every option that applied, so their figures represent the share of the 54 respondents choosing each option and sum to more than 100%. Open-ended responses were coded to recurring themes on review, and a response may be counted under more than one theme. All percentages are rounded to the nearest whole number, so totals may not sum to exactly 100%.
The sample is a self-selected group of asset owners rather than a random draw from the population, and it skews toward Americas-based organizations in the $1bn to $20bn range. Results describe the respondents themselves and are best read as indicative of engaged practitioners in this segment, rather than as an estimate projected to the wider universe of asset owners.