When AI picks who we fund

2 min read · July 18, 2025
New Power Labs

A 2022 study by Lyonnet and Stern found that a machine learning algorithm can outperform traditional VC decision-making.

While VC-backed firms tend to outperform the average, VCs still make costly mistakes, often backing firms that underdeliver. Eliminating the bottom half of a typical VC portfolio could increase returns by as much as 48%.

VC decisions are often swayed by founder demographics like age, gender, and education. 

In the study, the researchers built an algorithm that selects firms for investment based on characteristics that correlate with strong performance, measured by total sales and value-added figures from tax files. 

The research notes that by weighting founder demographics less heavily, the algorithm avoided firms likely to fail within five years and identified more “super performers” than VCs did.

This resulted in a nearly twofold increase in the proportion of female-founded firms, to approximately 18%, enhancing overall VC performance by 6%.

Still, as we know, AI is not a fix-all.

Recently, Workday is facing a class action lawsuit alleging its screening tool discriminates against job seekers based on race, age, and disability. In 2018, Amazon scrapped its AI hiring tool after discovering it penalized female applicants.

AI reflects the data it’s fed. If we train it on bias, it will scale bias. If we train it on equity, it can scale inclusion.

AI can help us move beyond the limits of human judgment, but only if we do the human work first.

Narinder

New Power Labs

Like what you’re reading? Subscribe to get weekly Equity Shots in your inbox.