AS THE AI GOLD RUSH BARRELS FORWARD, VCs EXPLAIN WHAT KIND OF AI STARTUPS THEY ACTUALLY WANT TO INVEST IN

The venture capitalist Gregg Hill spends a good amount of his working hours sussing out which companies are building AI products with staying power, versus those merely looking to cash in on hype. "The majority of companies are incorporating AI into their pitch decks," says Hill, a co-founder and managing partner of Parkway Ventures. "Every startup has dot AI" in their website url, he says, reflecting on the growth of the generative AI landscape. The technology that can generate text, video, audio, and more via written prompts has become entrenched in everyday business lingo and the cultural zeitgeist since OpenAI released ChatGPT in late 2022. It's kept investors busy wading through tremendous amounts of noise, as others make bets seemingly for the hell of it. "A lot of VCs, no offense to them, they're just getting into AI because it's the trend," Hill says. Though it now seems all-consuming, the AI revolution came just as another tech gold rush--characterized by web3 and crypto companies--was losing its luster. Startups and their financiers are now dealing with the aftermath of the industry's record funding boom, which began during the pandemic and wound down in 2022. Compared with the banner year of 2021, when total funding to U.S. startups reached $329 billion, things are much drier now: AI is one of a few categories netting investment anymore, alongside health care and biotech. As a result, startups are piling into the space and barraging investors with AI-enabled this and AI-powered that. Many investors have fallen into the FOMO trap, Nagraj Kashyap, general partner at Touring Capital, explains to Inc. Fear of missing out is spreading through the VC landscape, giving Kashyap flashbacks to the flusher days, when interest rates were rock bottom and money was free-flowing. "That whole era of companies getting priced at valuations that were not defendable ... And you would expect the investor community to learn from that. But we don't see that learning happening," he explains. It may seem like the AI hype train is destined to crash, as a few core giants such as OpenAI, Anthropic, and Cohere absorb funding from tech giants, and compete to build the best large language models (LLMs) in the game. But VCs who spoke to Inc. explain that there is massive promise for smaller players--and the majority of it lies in tools developed by companies that can cater to specific business needs. "Is there a real world problem that is not 'Hey, I can't generate text fast enough'?" asks Amias Gerety, a partner at the fintech investor QED. For founders, Gerety's question could prove instructive. Investors are looking to AI startups that offer solutions for businesses, or provide a modern makeover for technologies that haven't evolved in decades. A lot of the time, the applications are quite niche, even bordering on the mundane. Gerety points to Ocrolus, one of QED's portfolio companies, as an AI startup that has recognized and cornered a particular need: The company's tools extract data from pay stubs and bank statements to help lenders make better decisions. An AI tool for financial underwriting won't carry the mainstream resonance or shock value of a text-to-video tool like OpenAI's Sora. But less flashy, behind-the-scenes tech is what many VCs believe will eventually reshape the working world in AI's image. Kashyap mentions Netradyne, an AI-powered driving monitor that fleet vehicles use with the goal of fewer accidents and traffic violations in mind. "In real time, [Netradyne is] figuring out whether a truck is basically braking too hard, is almost tailgating, is going through stop signs without stopping," he says. As far as moonshot investments are concerned, Hill of Parkway isn't backing the makers of LLMs. But the ambition is still there: Parkway was a lead investor in the $23 million Series A fundraise of Oxos, a medical technology company building what it touts as a "radiology department in a box." Its device, a handheld X-ray machine called AiLARA, automates the appropriate amount of radiation necessary to achieve correct results. Anyone pitching VCs on an AI startup needs to consider how their tools can cater to a specific niche. "Is there a sales process that is broken and could be fixed with better payments? Is there a business interaction where you could make it easier for business X and business Y to coordinate?" Gerety asks, summarizing the kind of pointed questions founders should be asking. Of course, there is always the possibility of a bigger incumbent gobbling up whatever innovation a startup creates. Which is what makes the challenge so daunting. The ultimate sweet spot, Gerety says, is when "it's not just hard for your competitors to copy you, but your product gets better relative to the competition as you grow." The use of enterprise AI tools is already surging. A 2023 survey from workplace productivity tool Asana found that 36 percent of U.S. workers are already consulting AI tools at work at least once a week. The demand is there--and so is the money, if you've got the right specialty.