- Business Insider talked to a VC who's already using AI to replace associates.
- VCs are using AI more to scrape publicly available data to source hot companies.
- Still, VC will always be a human-to-human business where winning the best deals is most important.
When the longtime venture investor Matt Krna started Two Meter Capital this year, he hired only a skeleton team to manage a sprawling portfolio of more than 190 companies. The firm uses generative AI to do the bulk of its day-to-day portfolio management, tracking how companies perform and when they might raise another round. Without AI, Krna estimated he would have needed to hire half a dozen analysts.
"In terms of personnel efficiencies, if you were hiring an analyst class of six, now it's three," Krna said. "I think a lot of venture funds will start to use AI as a tool."
Silicon Valley has been enraptured by artificial intelligence for the past two years, with VCs racing to fund all manner of AI startups. Now, AI is changing venture capital itself, making what was already a hard field for young associates to break into that much more difficult and influencing how early-stage startups are funded.
Firms like Correlation Ventures, 645 Ventures, and Fly Ventures have long used data and AI to help guide investment decisions. SignalFire, a San Francisco firm that has been a leader in using data, developed an expensive platform years ago that tracks more than 10 million data sources. But rapid advances this year are making the use of AI more widespread, multiple venture capitalists told Business Insider.
Earlier this year, Sri Chandrasekar, a managing partner at Point72 Ventures, noticed a startup in his portfolio was having a breakout week.
"Because we're insiders, we saw, 'Wow, the company's doing really well,'" he said.
He thought he would have the data to himself, given that Point72 has access to the company's weekly internal metrics. But he was surprised to discover a flurry of interest from other firms, which he thinks can only be explained by competitors using AI models to comb publicly available data from thousands of companies.
"That's not an accident," Chandrasekar said. "There are signals that, if you know how to mine them on the internet, would identify if a company had a particularly good week, and miraculously seven funds reached out to them that week."
Chandrasekar declined to name the startup or the exact metrics that were so favorable but said they could include things like product utilization or frenetic hiring on the sales team. Firms that do not use AI to source deals will be left behind, Chandrasekar predicted. It will be up to general partners whether they want to use the technology to eliminate or enhance junior investing roles.
"Some of them are going to want better coverage with the same number of people, and some will just need fewer people," Chandrasekar said. "I'd rather get better coverage so I can look at more great deals."
Bain Capital Ventures, the venture division of Bain Capital with $160 billion under management, recently built a machine-learning model that helps identify inflecting companies the firm's partners should examine more closely.
Christina Melas-Kyriazi, a partner at Bain Capital Ventures who focuses on fintech and application software, said the model helped her identify a hot startup that was not on her radar because it was not in a tech hub.
"We noticed that this company had tremendous growth, and not only that but tremendous engagement from their users," Melas-Kyriazi said. "We immediately flew out to meet the founder and ended up making the investment. We wouldn't have otherwise known about this company if it wasn't for that model."
Back offices could be reduced by as much as 50%
Whatever happens to investing teams, Chandrasekar said he expected the size of back offices — handling tasks like human resources, administration, and financial reporting — to shrink more than 50%.
"If you go to any venture firm's website, you'll find that half the names are not doing anything to do with investing," Chandrasekar said. "I think the back office of every company is going to get impacted by AI significantly."
Andreessen Horowitz employs over 500 people, with its investment team growing 170% from 2017 to 2021. General Catalyst employs 259 people, while Lightspeed has 300, according to LinkedIn data.
"If you look at the large firms, they got very, very large in the last few years," Andy McLoughlin, a managing partner at Uncork Capital, said.
James Currier, a general partner at NFX, wrote a much-discussed column last year laying out why he thought the use of AI would level the playing field for venture investors in the next decade, just as software did for stocks and bonds in the past 40 years.
"Let's be honest: Much of what a typical venture capitalist does — reading, summarizing, and ranking — is what large language models already do extremely well," Currier wrote. "We are in the final 10 years of venture capital as we have come to know it. A.I. is going to remake the startup industrial complex, from its core. Venture firms will have to remake themselves into a combination of people and A.I."
Still a human-to-human business
Overall, it is not as if most VC firms are brimming with staff. The average firm employs just 14 people, according to Deloitte.
"They're already so small," a partner at a large firm who asked to not to be identified said, speaking in a vast conference room in their sleek and mostly empty San Francisco office. "Could you have fewer people? To what end? You're making so much money as it is."
And central to making that money is charging rich management fees for the mystique of venture investing, which most VCs insist is mostly about gut feeling, intuition, and personal connections that AI can never replace.
"It's still very much a human business," McLoughlin said. "People want to work with people."
Uncork employs only one associate, and McLoughlin said he would not replace her with AI anytime soon.
"I'm sure there's a bunch of AI tools that she's using, but her value is in being very, very smart and extremely well trained," he said.
Humans are particularly hard to replace at the earliest stages, where it's often more about the kernel of an idea with little data for AI to harvest.
"I don't think AI is very good at analyzing things we haven't seen before, and the best outcomes in venture often break the pattern," Melas-Kyriazi said.
At its heart, venture is not about finding the best deals but winning them. That comes down to a personal connection between founder and investor, Chandrasekar said.
"The No. 1 reason a founder chooses you is they want to work with you, specifically," he said. "You spend time building a relationship with them. You have dinner with them. You meet their family — whatever it is you need to do to make them feel like they really want to work with you for what you hope is a 10-year marriage."
Machines are a long way from that.
