Large limited partners have moved past curiosity about AI. Public pension funds are among the most deliberate technology buyers in private markets. When one of them needs a new data platform, document management system or research provider, the purchase goes through a formal request for proposal, often with sign-off from an investment committee or board.
Major limited partners' 2026 technology and data procurements give us a window into how they are thinking about AI. Most are asking detailed questions about how it works and how it is controlled. They are mostly considering it for middle and back office use cases and judging whether it will make them faster, more accurate and more dependable. We expect these funds to bring the same questions to their managers, and smaller limited partners to follow their lead.
The examples below come from RFPs and requests for information that large US public pension funds, along with a Texas state trust company that manages endowment money, issued over the past year. They cover software, data platforms and research services. Only one is an AI project on its face, yet every one of them asks vendors about AI. Read together, they show that in 2026 AI sits at the center of how these funds choose vendors.
AI Now Has Its Own Line in the Evaluation
A year or two ago, a question about AI might have shown up near the end of a vendor questionnaire, if at all. Now it often gets a section of its own.
Massachusetts PRIM gave AI a dedicated "Technology and Innovation — Artificial Intelligence" section in its search for a proxy research provider.¹ Bidders for the $129 billion fund had to explain where AI sits in their workflows, how it affects staffing on the MassPRIM account, and how the guardrails around it are "monitored and updated over time."
At the $168 billion Texas Treasury Safekeeping Trust Company (TTSTC), AI was a required response item in the search for an investment data platform.² Ohio PERS, with more than $120 billion, built AI questions into the standard template it uses across solicitations, so every vendor that bids now has to explain the models behind its product and how the fund's data is handled.³
Indiana's INPRS went a step further and sought advice on building AI as a program in its own right. This year the $55 billion fund sought advisers to build an AI governance framework and a 12-to-24-month roadmap, with the stated aim of moving "from AI enablement to disciplined execution."⁴
Efficiency Is the First Ask
So what do these funds want AI to do? Mostly, they want it to take on the repetitive work that slows their operations teams down.
Colorado PERA, at $75 billion, set an explicit goal in its investment systems modernization to use AI "to improve productivity, reduce operational risk, and enhance analytical insights and reporting."⁵
The New Jersey Division of Investment, whose $86 billion pension fund holds about a quarter of its assets in private markets, wants a document platform that pulls capital calls, distributions and financial statements out of email and classifies them. Bidders must explain which steps the system handles with AI and which still need a person.⁶ In its search for a knowledge management platform, the $94 billion LACERA asked for generative AI that can summarize large sets of documents and answer staff questions.⁷
Quality and Reliability Carry Equal Weight
Faster is only useful if the output is right. Anyone who has reconciled a capital account knows that a quick wrong answer costs more than a slow correct one, and these funds write that concern straight into their RFPs. TTSTC wants "Machine Learning assisted anomaly detection" in the data quality layer of its new platform, which will hold a reconciled golden copy of its investment data.² MassPRIM asked vendors to describe their AI guardrails "with respect to accuracy, bias, data quality, and the independence of research,"¹ and Minnesota's $171 billion State Board of Investment used close to the same language, asking for "associated controls for accuracy, bias, and analytical independence."⁸
Reliability shows up as control. TTSTC's contract terms require vendors to run "model integrity validation checks," monitor for data drift and give the state the option to turn the AI off.² Ohio PERS wants to know whether the fund's data trains the vendor's model and who can see it,³ while MassPRIM asks for notice of material changes to a vendor's "AI approach" during the contract.¹ INPRS has written human-in-the-loop design into the requirements for its roadmap.⁴
Managers Are Next, and Smaller Limited Partners Will Follow
For technology and data vendors, a credible AI answer is now part of winning institutional business. It should name the models, show where human review fits and demonstrate accuracy on the client's own documents.
We expect general partners to face the same questions soon. A fund that asks its data vendor how AI catches errors will reasonably ask its managers how they produce capital account statements, valuations and portfolio reports, and where AI sits in that process. The questions will start in operational due diligence and in due diligence questionnaires.
Over time they will extend to what limited partners expect to receive, which is cleaner data delivered faster in formats their own systems can take in without rekeying. A manager whose reporting still depends on manual spreadsheet work will find it harder to defend once its largest investors have automated their side of the exchange.
The shift will spread beyond the largest funds. Smaller pensions, endowments and foundations tend to adopt the diligence frameworks that larger peers and their consultants put in place, and AI questions will travel the same route. The vendors serving these funds will build AI features to win the largest mandates and then roll them out to every client, which will raise expectations across the market. Within a few years, a mid-sized limited partner may ask its managers questions about AI that only the biggest pensions ask today.
Our advice to managers is to get ahead of these questions. Put AI to work on one operational problem, such as investor reporting or document intake, and be ready to show results from work already in production.
Footnotes
¹ Request for Proposals: Proxy Advisory and Voting Services, Massachusetts Pension Reserves Investment Management Board, May 2026. Assets per PRIM Performance Summary, July 31, 2026.
² RFI No. 311-26-180526AK: Investment Data Platform Solution, Texas Treasury Safekeeping Trust Company, May 2026. Assets per Trust Company AUM Report, August 31, 2025.
³ Request for Proposal: Proxy Services & Voting Platform, Ohio Public Employees Retirement System, July 2026. Assets per OPERS Defined Benefit Fund reporting, June 30, 2026.
⁴ RFP 26-02: AI Roadmap Development, Indiana Public Retirement System, March 2026. Assets per INPRS Annual Report, Fiscal Year 2025.
⁵ Request for Proposals: Investment System Modernization – Roadmap Development and Advisory Services, Colorado PERA, June 2026. Assets per PERA 2025 Annual Report.
⁶ Request for Quotes: Documentation Management and Workflow Solution, New Jersey Division of Investment, June 2026. Assets per Director's Report, May 2026.
⁷ Request for Proposal: Knowledge Management Solution, Los Angeles County Employees Retirement Association, October 2025. Assets per LACERA Total Fund Performance Report, Q2 2026.
⁸ Request for Proposal: Proxy Voting Services, Minnesota State Board of Investment, July 2026. Assets per SBI, June 30, 2026.
