Some VC firms are using AI algorithms to help with investment decisions; Gartner forecasts that AI will be involved in 75% of VC investment decisions by 2025
Jared Council / Wall Street Journal : Tweets: @trengriffin , @ryanlawler , and @hlandgren Tweets: Tren Griffin / @trengriffin : Everyone knows that AI will be involved in 76.4% of venture capital decisions in 2O25, not just 75%. One AI in VC use case is keeping track of whether the financing is a Series P or a Series Q round, which is always tricky. https://twitter.com/... Ryan Lawler / @ryanlawler : This story pops up every couple of years and I've seen v. little evidence that the data collected and analyzed is actually leading to better investment outcomes https://www.wsj.com/... Henrik Landgren / @hlandgren : Great article in @WSJ about how investors are adopting more and more AI to make better decisions. We've come along way already (9 investments, 100MUSD+ invested, first exit and first unicorn) but its really just the beginning. https://www.wsj.com/...
Context & Ripple Effects
This 2021 WSJ piece was an early marker in a now well-documented arc: Gartner's forecast that AI would touch 75% of VC investment decisions read as speculative at the time, and voices like Ryan Lawler were already flagging that the underlying data rarely improved actual returns. By 2023 the practice had gone mainstream enough that KPMG, Coatue, and Headline were openly using AI to screen deals and acquisition targets.
What followed validated both sides of the debate. The dollars concentrated dramatically — AI startups took roughly 30% of VC money in 2024 per one analysis, then 53% globally in H1 2025 — and by Q1 2026 a record quarter saw AI capture 81% of a $297B total. The question shifted from whether VCs use AI to whether the machines will soon do more than assist.
First-order effects
- Firms adopting AI deal-screening gain speed and breadth in sourcing, forcing competitors like Coatue and Headline — already public about their tooling — to treat proprietary data pipelines rather than partner networks as the scarce asset.
- Founders pitching funds now face a first filter that may be an algorithm, raising the bar for metrics-clean decks before any human partner weighs in.
Second-order effects
- If every firm runs similar models on similar market data, the edge compresses — pushing VCs toward exclusive data sources, internal operating metrics, or the deeper automation of analysis described in later coverage of agentic investors reviewing pitch decks.
- LPs get a new diligence question: whether algorithm-assisted selection actually improves fund returns, a gap skeptics like Lawler have argued remains unproven even as AI's share of deployed capital keeps climbing.
Third-order effects
- The pattern points toward venture bifurcating: a commoditized algorithmic screening layer beneath a shrinking set of conviction bets, with the sector's own concentration (AI taking 81% of record Q1 2026 dollars) meaning the models are increasingly trained on, and allocating to, AI companies themselves — a feedback loop whose stability is genuinely untested.
- Regulatory and LP scrutiny of automated capital allocation becomes plausible once 'AI was involved' stops being an assist and starts being the decider, especially given the forecast's own framing made involvement near-universal by design.
The trend: Venture capital is moving along the path this article charted — from AI-assisted screening in 2021 toward AI as primary analyst — just as the industry concentrates most of its capital into AI companies, tightening the loop between allocator and allocated.