Everyone's Thesis but Their Own
Family offices have made artificial intelligence their highest-conviction investment. They just won't use it.
There is an investor, a composite drawn from three major surveys published in the past twelve months, who has committed a meaningful portion of a multi-billion-dollar portfolio to companies building artificial intelligence. She has exposure through public equities, through venture allocations to frontier AI labs, and through a co-investment in a semiconductor company designing the chips that will power the next generation of large language models. She ranks AI as her top conviction theme for the next five years. She has told Goldman Sachs as much.
She has also spent the last three weeks trying to get her family office's Q4 report finished. The process involves downloading statements from six custodians, reconciling them in Excel, and manually building a PDF that her principal will skim for twelve minutes before asking a question that requires another three days of data retrieval to answer. She does this every quarter. She has been doing it for years.
Nobody in the office has suggested that the technology they are betting the portfolio on might also be useful for running the institution that manages it.
This is the adoption paradox, and in 2026, it has become the most revealing contradiction in the family office sector. The paradox is not new. What is new is that the data documenting it has finally become impossible to ignore.
The numbers that don't add up
Begin with what family offices believe. Goldman Sachs' 2025 Family Office Investment Insights report, drawn from 245 institutional family offices globally, found that 86 percent now invest in AI in some capacity. More than half, 51 percent, already use AI in their investment decision-making processes. A full 58 percent expect their portfolios to be overweight technology in the next twelve months. AI is not a speculative interest for this cohort. It is the consensus trade.
BNY Wealth's 2025 survey echoed the conviction: 83 percent of family offices rank artificial intelligence among their top five investment themes for the next five years. Bank of America's inaugural Family Office Study, surveying 335 decision-makers across North America, found that 57 percent have already utilized AI for investment research and strategy, with automation widely adopted for forecasting (76 percent), alternative investment analysis (74 percent), and portfolio modeling (73 percent).
Now set those numbers beside what family offices actually do inside their own walls.
The J.P. Morgan Private Bank's 2026 Global Family Office Report, surveying 333 single family offices across 30 countries with an average net worth of $1.6 billion, found that while 65 percent intend to prioritize AI, over 70 percent currently have zero investment in the digital infrastructure that would make any of it work. BlackRock's 2025 Global Family Office Report, covering 175 single family offices with average assets of $2 billion, was more direct: only one in three family offices currently use AI internally. Citi Private Bank's 2024 Global Family Office Survey reported that more than 53 percent of family offices globally have invested in generative AI, yet fewer than 15 percent report using it for practical applications such as automation, forecasting, or reporting.
That gap, between 86 percent investing in AI as an asset class and roughly 15 to 33 percent deploying it as an operational tool, is not a minor inconsistency. It is a structural contradiction. These institutions have identified the technology that they believe will reshape the global economy, and they are funding it with extraordinary conviction. They are then walking past it every morning on the way to a spreadsheet.
The paradox, explained
The instinct is to attribute this to ignorance or inertia, but neither explanation survives contact with the evidence. Family office leaders are not unaware that AI exists as an operational tool. They are, by the surveys' own admission, already experimenting with it in investment analytics. The disconnect is more specific, and more interesting, than a simple failure to pay attention.
Four forces sustain the paradox.
The first is a categorical error in how family offices classify AI. In the investment committee, artificial intelligence is an asset class: a sector allocation, a thematic bet, a source of future returns. In the operations room, it is a technology purchase: a cost, a vendor relationship, a disruption to existing workflows. The same technology is evaluated under two entirely different frameworks within the same institution. In one room it is opportunity; in the other, overhead.
The second is the privacy instinct. Family offices exist, in part, to protect the most sensitive financial, legal, and personal information a family possesses. The idea of routing that information through a cloud-based AI platform, even one with enterprise-grade encryption, triggers a reflexive caution that is understandable, even if it is sometimes disproportionate. As one director put it in BlackRock's survey, the desire to adopt AI is there, but the clarity about how to do it safely is not. The fear centers on what happens when discretion meets the cloud.
The third is the lean team problem. The Goldman Sachs report notes that family office investment teams are typically made up of fewer than five individuals. When your entire operation runs on a handful of people, the switching cost of adopting any new system is measured in bandwidth, and bandwidth is the scarcest resource a family office has. The APQC ratio resurfaces here with new implications: if 75 percent of your team's time is already consumed by data gathering, nobody has the remaining hours to evaluate, implement, and learn a new technology platform. The operational drag does not merely waste time. It prevents the institution from investing in the very tools that would stop the waste.
The fourth, and perhaps most fundamental, is the absence of an internal champion. In a family office, the principal sets the investment thesis. The CIO executes it. But who owns the operational technology agenda? In many offices, the answer is nobody, or, more precisely, whoever happens to be most frustrated with the current process. There is no chief technology officer. There is no digital transformation mandate. There is a CFO or COO who also handles compliance, insurance, and facilities, and who does not have the time or the organizational authority to propose a systemic overhaul of how the institution processes information.
The cost of the contradiction
Articulating why the paradox persists is useful. Quantifying what it costs is more so.
Return to the staffing math. A family office employing five investment and operations professionals, the Goldman Sachs median, and losing 75 percent of their time to data administration is operating with the equivalent of 1.25 full-time analysts doing the work that justifies the office's existence. The other 3.75 person-equivalents are occupied with tasks that a properly deployed AI system could reduce by half or more. This is not speculation: it is what the offices that have already done it report. One family office quoted in BlackRock's report described replacing three days of monthly Excel work with a thirty-second automated script.
Three days per month is thirty-six days per year. For a single employee, that is roughly 15 percent of their total working time recovered. Applied across a five-person team, the gains are structural.
IBM's 2024 Cost of a Data Breach report found that organizations using AI and automation in their security operations saved an average of $2.2 million per breach compared to those that did not. In financial services, where the average breach cost is $6.08 million, more than 20 percent above the global mean, that savings is the difference between a survivable incident and an existential one.
And then there is the opportunity cost that never appears on any ledger. Every hour spent reconciling custodian statements is an hour not spent evaluating a co-investment. Every week consumed by quarterly report assembly is a week in which a market dislocation goes unexploited. McKinsey found that 41 percent of CFOs report that a quarter or less of their processes are currently digitized or automated. In family offices, where technology budgets are a fraction of institutional equivalents, the figure is almost certainly worse. The paradox is not just that family offices are not using AI internally. It is that the absence of AI internally is degrading the quality of the very investment decisions that AI as an asset class is supposed to enhance.
The investors who are most bullish on artificial intelligence's capacity to transform every industry are running the one institution they could transform tomorrow on a system that would embarrass a mid-market accounting firm.
The offices that moved
The story is not entirely static. RBC and Campden Wealth's 2025 report found that three times more family offices are leveraging AI to improve operations compared to the prior year. Bank of America's survey revealed that nine out of ten family offices believe AI could enhance investment returns, and half have already experimented with it in some capacity. Simple's 2025 Family Office Software & Technology Report noted that fewer than 25 percent of new prospects now rely on spreadsheets as their primary tool, a meaningful decline from the historical norm.
These are early signals rather than a transformation, but they suggest where the movement is coming from. The offices making progress share three characteristics.
They start with pain points, not platforms. The most successful early adopters are not attempting to overhaul their entire technology stack at once. They are identifying the single most time-consuming manual process, typically data aggregation or report generation, and automating that one workflow. Trust agreement summarization, where AI extracts key provisions from complex legal documents, has emerged as a particularly high-value, low-risk starting point that PwC has highlighted as a standout use case for family offices.
They appoint an internal owner. The offices that stall are the ones where AI adoption is everyone's interest and nobody's responsibility. The ones that succeed have designated a single individual, sometimes formally, sometimes informally, to coordinate AI evaluation, manage pilots, and translate results to principals. The role requires someone with enough institutional credibility to bridge the gap between the investment committee's enthusiasm and the operations team's skepticism, though it does not require a technologist.
They govern before they deploy. Privacy and security concerns are legitimate and should be taken seriously. The offices moving fastest are the ones that have established clear governance frameworks before deploying any AI tool: defining what data can and cannot be processed, who reviews AI-generated outputs, and how accountability is maintained. Governance is the precondition for adoption that endures.
The twelve-month window
There is a timing dimension to this paradox that sharpens the stakes. AI capabilities are not developing on a schedule that accommodates institutional deliberation. The technology available today is materially more powerful than what existed twelve months ago, and what will be available twelve months from now will make today's tools look primitive. Hyperscaler capital expenditure on AI infrastructure is forecast to exceed $600 billion in 2026, a 36 percent increase year over year. The models, the platforms, and the applications are coming whether family offices are ready for them or not.
The generational factor accelerates this further. Bank of America found that 87 percent of family offices have not yet been passed to the next generation, and 59 percent expect that transition within the next decade. JPMorgan Private Bank's research found that nearly 80 percent of ultra-high-net-worth principals already use AI in their personal lives, and 69 percent use it within their businesses. The next-generation leaders inheriting these institutions will not arrive asking whether AI should be adopted. They will arrive asking why it hasn't been.
The adoption paradox, in other words, has an expiration date. It will be resolved either by the offices that choose to close it deliberately, starting now, starting small, starting with governance, or by the generational turnover that will close it for them. The former is a strategic advantage. The latter is a scramble.
For an industry that prides itself on long-term thinking, the irony should be uncomfortable. The most patient capital in the world is being remarkably impatient about AI when it appears on a term sheet and remarkably patient about it when it could be running their operations. Everyone's thesis. Nobody's implementation.
The gap between conviction and conversion has never been wider, and the window to close it on your own terms has never been shorter.
This is the fourth installment of The Prominent Dispatch, a biweekly series on the convergence of capital strategy and operational technology in the family office sector.
Sources and verification notes:
All statistics cited in this article are drawn from identified, published sources:
Goldman Sachs 2025 Family Office Investment Insights Report ("Adapting to the Terrain"): 245 institutional family offices surveyed globally (May–June 2025); 86% invest in AI; 51% use AI in investment processes; 58% expect to overweight technology in next 12 months; investment teams typically fewer than 5 individuals. Press release September 10, 2025; confirmed via Goldman Sachs, CNBC, MarketScreener, Worth.
BNY Wealth 2025 Global Single Family Office Survey: 83% rank AI among top five conviction themes for next five years. Confirmed via BNY Wealth website (November 2025).
Bank of America Family Office Study 2025 ("Perspectives on the Modern Family Office"): 335 decision-makers across North America surveyed (May–June 2025); 57% use AI for investment research and strategy; automation used for forecasting (76%), alternative investment analysis (74%), portfolio modeling (73%); 90% believe AI could enhance returns; 87% not yet transitioned to next generation; 59% expect generational handover within 10 years; nearly 1/3 experienced cyberattack. Press release November 12, 2025; confirmed via Bank of America, PR Newswire, Nasdaq, InvestmentNews, Morningstar.
J.P. Morgan Private Bank 2026 Global Family Office Report: 333 SFOs, 30 countries, avg net worth $1.6B; 65% intend to prioritize AI; 70%+ have zero digital infrastructure investment. Confirmed via J.P. Morgan and MyFO Tech analysis (February 2026).
J.P. Morgan Private Bank, "Stewardship & Purpose: Conversations with the World's Wealthiest Families" (2025 edition): Nearly 80% of UHNW principals use AI personally; 69% use AI within their businesses. Confirmed via Spear's (November 5, 2025), which references the report directly. Note: this is a separate publication from the 2026 Global Family Office Report.
BlackRock 2025 Global Family Office Report: 175 SFOs, avg AUM $2B; only 1 in 3 use AI internally; 51% invest in AI beneficiary companies; 45% back AI infrastructure enablers; 34% use AI for investment analytics. Confirmed via Investment Officer, Aleta.io (September–October 2025).
Citi Private Bank Global Family Office Survey 2024: 338 FOs surveyed (June–July 2024); 53% have built portfolio exposure to generative AI. Operational AI use at 12–13% per individual use case (automation 13%, presentations 13%, forecasting 12%); characterized as "fewer than 15%" by Crain Currency (March 2025) and WealthBriefing (2025). Primary survey confirmed via Citi Private Bank PDF and Caproasia summary.
Deloitte Private, Family Office Insights Series: "Digital Transformation of Family Office Operations" (2024): 354 SFOs surveyed globally plus 40 in-depth interviews; 12% using AI-driven solutions operationally. Confirmed directly from Deloitte Global and Deloitte UK report pages, and Family Wealth Report coverage.
Hyperscaler capex forecast: CreditSights projected ~$602B for top-5 hyperscalers in 2026, +36% YoY (November 2025 estimate). Confirmed by IEEE ComSoc Technology Blog (December 2025), CreditSights, MUFG Americas. Note: CreditSights revised estimate upward to ~$750B in February 2026 following Q4 2025 earnings calls; "$600B+" remains accurate as a conservative floor.
RBC and Campden Wealth, North America Family Office Report 2025: 3x increase in AI adoption for operations YoY. Confirmed via Wealth Solutions Report (October 2025).
Simple 2025 Family Office Software & Technology Report: 92% of platforms have live or developing AI capabilities; fewer than 25% of new prospects rely on spreadsheets as primary tool. Published November 2025.
PwC Family Office panel at 11th Family Wealth Report Summit 2025: Trust agreement summarization identified as standout AI use case. Reported by Family Wealth Report.
APQC via CFO.com: 75/25 ratio for FP&A time allocation (data gathering vs. value-added analysis). Previously verified in Article 2.
IBM Cost of a Data Breach 2024: $4.88M global average; $6.08M financial services; AI/automation users saved avg $2.2M per breach. Previously verified in Article 3.
McKinsey 2024: 41% of CFOs report 25% or less of processes digitized/automated. Previously verified in Article 2.
