AI will have its say in every trend and challenge on private equity’s (PE) plates in 2026: from heightened demands for transparent, precise reporting by limited partners (LPs) to a premium on creative deal structures, new bank-private lender dynamics, and a stable but uncertain exit environment. For some firms, AI is already changing the game, the field, and the teams, as a multiplier of automation and efficiency. Early adopters are transitioning from piloting a few use cases with generative AI to rewiring entire operations with a platoon of AI agents.
But many firms have been frustrated with the disappointing benefit curve of their initiatives as they integrate AI into workflows. Regardless, nobody can afford to fall behind in this technological race. Getting this early phase of AI transformation correct will go a long way in positioning private market firms to leverage this once-in-a-generation technology to achieve operational alpha and ultimately generate exceptional returns — if adopted with thoughtful AI governance and the right data foundation. There are a few fundamentals that can help point firms in the right direction in an adoption journey in which the questions outnumber answers.
What do PE’s want from AI?
EY recently found that 38% of PE firms are expecting to spend more than half of their total budget on AI in 2026, while 42% are currently directing at least 25% of their total business units’ budget toward AI.[i] The investment industry is moving beyond the first phase of the generative AI revolution. But what may not be obvious is what private market firms precisely need and expect from AI.
Too often large language model (LLM) copilots end up just amounting to bad interns churning out polished but error-ridden work.
Firms still need to step away from the time-consuming legacy workflows, since many are still married to manual processes. They need to make better, more informed decisions. They need to move opportunities faster and with greater confidence. Firms need to scale AUM without operational friction and added expenditures. They need the ability to generate real insights using AI tools to analyze data. Finally, they need their AI implementations to avoid becoming money pits.
AI magnifies operational weaknesses
As private market firms seek to progress their use cases toward AI solutions and AI agents that execute more advanced analyses and operational workflows, they are running into persistent errors and breaks, as well as trouble keeping their data ingestible, reportable, auditable, and accessible to teams who need it. Too often large language model (LLM) copilots end up just amounting to bad interns churning out polished but error-ridden work. Worse, AI can be a multiplier of operational weaknesses and risk exposure if implemented on inadequate, old data infrastructure.
Modern data foundation derisks agentic AI initiatives
If built on a flexible, centralized, automated data foundation, firms can confidently deploy AI agents to execute operations like reconciliations, processing of notices, or drafting investor reporting — tasks that would otherwise be painful, if not complete non-starters. The data foundation helps analysts do higher-value tasks because they can access data from multiple applications and proprietary storage formats. For example, an analyst can marry reference data to alternative market datasets and historical asset prices to run different types of exposure and risk computations.
A prerequisite is the system’s ability to map the structure and relationships between all of a firm’s datasets and pipes — what we refer to as data ontology.
Built on top of modern data infrastructure, AI agents can be deployed to swiftly build models and data pipelines to bring in new datasets, as well as help with data quality management. Agents can help analyze and fix reconciliation breaks. Analysts can use natural language AI tools to interrogate datasets and build custom reports and analyses, like traditional market and fundamental analysis, and exposure analysis and scenario testing. These time-saving operational tasks require cross-referencing position data, investment reference data, and historical resolution data, so attempting to layer AI over a fragmented, obsolete data foundation can lead to inaccurate outputs.
Move faster in dealmaking by mastering the data
General partners (GPs) love to move faster and more accurately than the next shop when it comes to reacting to market fluctuations and deal origination. For example, a manager at a private credit fund has been running a sidecar with one large LP on a deal‑by‑deal basis. The LP suddenly decides that it will only continue if the co‑investments are migrated into a formal commingled vehicle immediately, and it offers a sizable seed commitment on the condition that the PM can execute the first deal within days because of time‑sensitive financing. AI will be an essential piece of technology to do just that. How? The cloud revolution and general digitization helped the investment industry create a data monster on orders of magnitude too large to grasp. AI can process and extract information from large, often complex datasets at a scale and speed not previously possible. However, a prerequisite is the system’s ability to map the structure and relationships between all of a firm’s datasets and pipes — what we refer to as data ontology.[ii]
A configurable, interoperable data platform built for financial services firms will maintain that data ontology, or scaffolding, for the firm. AI can not only help to maintain and oversee the data scaffolding, but it will also do its higher value work on top of that scaffolding. Only then can a portfolio manager move swiftly to seize the opportunities that cross their desk.
AI decodes the hieroglyphics of unstructured data in private credit
Unlike their public market counterparts, private markets investments are awash with unstructured datasets from spreadsheets, emails, PDFs, and seemingly infinite loan tapes in asset-based finance. Private credit datasets are often built on undefined schemas, inconsistent identifiers, and unpredictable tabular formats — a nightmare for legacy systems. AI-powered unstructured data processing can extract loan lifecycle events like drawdowns, paydowns, and interest repricing from loan notices. This automation eliminates hours of manual validation by investment operations teams so they can focus on reviewing exceptions. Moreover, processing unstructured datasets from PE portfolio company financials, LP documents, and bilateral trading agreements is a tailor-made task for AI’s talents.
GPs send LPs quarterly, weekly, and sometimes daily communications for items like capital calls on unstructured products in the form of security information and transaction information. If a firm cannot smoothly and accurately ingest and transform unstructured data as well as it does structured data, humans will be re-enlisted to slog through onerous manual processes.
PE’s AI problems are data problems
No matter the dollar signs they throw at AI initiatives, private market firms’ AI aspirations depend on trusted data foundations to deliver value safely and without costly bottlenecks. AI-ready data infrastructure will efficiently organize structured and unstructured data, consolidate it across systems, and ensure it is interoperable and governed. The AI transformation is a marathon and not a sprint, and the winners will be cautious, thoughtful, and meticulous integrating this technology.
About the Author
Bryan Dougherty is President, Product and Technology at Arcesium. In this position he oversees platform development, infrastructure and security. Prior to the formation of Arcesium, Bryan was Head of Middle and Back Office Technology at the D. E. Shaw group for nine years. He began his career at Random Walk Computing, a technology consultancy focused on the capital markets where he worked on a variety of projects for exchanges/marketplaces, investment banks, and hedge funds.
Arcesium provides financial technology and data management services to private markets firms, hedge funds, and institutional investors. Its cloud-based platform handles complex workflows and centralizes data across the investment lifecycle. The company supports more than $5.3 trillion in AUM and $1.2 trillion in sell-side balances, and has modeled over 160 million investments. Arcesium was developed by D. E. Shaw Group and launched with Blackstone Multi-Asset Investing, with J.P. Morgan later investing. It employs more than 2,300 professionals.