As generative AI technology continues to gain traction, established businesses are adopting it at a rapid pace. That has left many startups, which normally lead the tech adoption curve, scrambling to keep up.
Venture capitalists are on a perennial hunt for the next groundbreaking technology. Generative AI may very well be it, and as in any hype cycle, it’s easy to back the wrong companies. But for early-stage VCs, investing in AI could turn out to be more challenging than in the previous tech revolutions.
The first major problem for VCs is that incumbents are already pouncing on the technology, either by building their own large language models or spinning up applications built on GPT-4 and its ilk. Boardrooms are buzzing with discussions of whether and how to add generative AI capabilities to their offerings.
“Most of the time, startups’ advantage has been they learn new technologies faster. But this time, they’re behind,” said Tomasz Tunguz, a well-known SaaS blogger and investor who recently left Redpoint Ventures to launch his firm, Theory Ventures.
Agile incumbents
It’s not only pure tech companies that are eager to add this technology. Bloomberg has developed a proprietary large language model trained on its vast wealth of financial data. Meanwhile, 12-year-old Bloomberg competitor AlphaSense received $100 million for building generative AI features from Alphabet‘s late-stage corporate venture arm CapitalG.
This article appeared as part of The Weekend Pitch newsletter. Subscribe to the newsletter here.
Morgan Stanley is another financial services incumbent rushing to incorporate this tech by working with OpenAI on a chatbot for financial advisors. Other financial institutions likely aren’t far behind.
Similar capabilities will soon be integrated into other tech-enabled companies across various industries from healthcare and education to insurance and travel.
“This is the conundrum: it’s a game-changing technology, but the question is how does the startup take advantage of it relative to an incumbent? It’s not clear,” Tunguz said.
Huge niches
Another issue is that many founders of new generative AI startups are not thinking through how to make their businesses differentiated and defensible.
James Currier, co-founder of NFX, a seed-stage focused VC firm, told me in December that he was disappointed with all the startups he saw in the space then.
A lot of the use cases in generative AI “are very obvious,” he said. That’s why NFX created a market map of the field—to show founders that somebody else is probably already working on that idea.
When I talked to Currier this week, he sounded more optimistic. While he still has not discovered “visionary” applications of these technologies, such as when the mobile tech wave gave rise to marketplaces like Airbnb and Uber, he is now seeing companies that have figured out how to create defensible businesses.
He didn’t go into specifics other than tell me that until those visionary ideas materialize, founders should go for the niches without clear incumbents, and areas where giants like OpenAI in partnership with Microsoft or Google wouldn’t want to participate.
One such area is law. Legal firms aren’t likely to create this tech for themselves, and generative AI could be highly effective at picking out salient aspects of dense documents. Currier touted NFX investments in Darrow, a startup that searches public records to identify potential class action lawsuits, and Evenup, an AI-powered software for personal injury lawyers that recently received a $350 million Series B valuation from Bessemer at a steep 40x revenue multiple, The Information reported.
“Personal injury [law]—you can build a $5 or a $10 billion company on that,” Currier said. “It’s a huge industry that’s been ignored.”
Another notable startup trying to take advantage of legal offerings is Harvey, an AI assistant for lawyers, which just raised a Series A at a $150 million valuation led by Sequoia, Insider reported.
While some niche areas have not been served well by technology until recently, and investors encourage entrepreneurs to look for these under-the-radar white spaces, VCs urge building as fast as possible once the opportunity is identified.
In fact, speed has never been more paramount.
“Everything is moving so fast,” said Tunguz. “In the [early] internet [era] 20 years ago, you probably had 18 months lead time. If you built a mobile app, you had 12 months. Now, with generative AI, you have a couple of weeks to a couple of months at best.”
Featured image by Mara Potter/PitchBook News
Learn more about our editorial standards.
