Private markets GPs accelerate AI adoption across deal origination and diligence

Private markets managers are pushing artificial intelligence deeper into the deal funnel, with the clearest gains concentrated in sourcing, screening, and financial modeling at the general partner level, according to a report published by Accordion on Monday.

The report describes GPs using AI to scour public filings and news articles for deal leads, draft targeted pitch letters to acquisition prospects, and produce offer letters. Quality-of-earnings work and the marking of portfolio assets to market are also moving into AI-assisted workflows, per Accordion.

Hebbia, an AI platform built for financial services users, supplies one of the named tools in the workflow. Its virtual data room product ingests deal files and screens opportunities against specific investment criteria, founder and CEO George Sivulka told Accordion. The pitch is throughput: filtering opportunities against a fund's mandate in a fraction of the time a human associate would need to read the same documents.

The modeling use case is where the report puts a number on the productivity claim. An analyst at an unnamed private credit firm built a full CLO-to-cash-flow model from scratch in two hours using Anthropic's Claude Opus, a task that previously took materially longer, Accordion reported. CLO modeling is a defined, structured exercise: tranche waterfalls, coupon schedules, default assumptions, and recovery curves. It is exactly the kind of work where a frontier LLM with good context handling can compress a junior analyst's week.

What's actually shipping at the GP level

Accordion groups the early GP use cases into three buckets:

• Sourcing. Filings and news screening for deal leads; auto-drafting of pitch letters with prospect-specific rationale; offer letter generation.
• Diligence. Virtual data room ingestion and risk screening against investment criteria, via tools like Hebbia.
• Modeling and portfolio work. Quality-of-earnings analysis, financial models built from primary documents, mark-to-market on portfolio assets.

Notably absent from the list: anything that touches limited partner reporting, fund-level performance attribution, or LP-side workflows. The report's framing is that AI gains are concentrated upstream, in the work that gets a deal from pipeline to close, rather than downstream in fund operations.

Why adoption is uneven

The Accordion piece flags that pace varies. It does not name the laggards, and it does not quantify how many managers are still in pilot mode versus production. That gap matters. "Stepping up" can describe a fund that has rolled Hebbia into every deal team's diligence stack, or it can describe a fund that has bought ChatGPT Enterprise seats and called it a strategy.

For expert networks and research providers serving the buy side, the more useful question is which side of the spread a given GP sits on. A fund where associates are building CLO models in Claude in two hours has a fundamentally different research budget than one still routing every modeling task through a junior team.

The report did not disclose which private credit firm ran the CLO modeling exercise, nor did it specify the size of the deal team or the broader Claude deployment at that fund. Those details would sharpen the picture of how widely the two-hour modeling claim generalizes across the private credit cohort.