AI companies worldwide will raise more than $100 billion dollars in venture capital by 2024, according to Crunchbase dataan increase of more than 80% compared to 2023. This will make up nearly a third of the total VC dollars invested in 2024. That’s a lot of money funneling to many companies in AI.

The AI ​​industry has grown so much in the last two years that it’s filled with overlapping companies, startups that still use AI in marketing, but not in practice, and legitimate diamonds-in-the-rough. which are AI startups that are grinding. Investors have their work cut out for them when it comes to finding startups that have the potential to become category leaders. Where do they start?

TechCrunch recently surveyed 20 VCs who backs building startups for businesses about what gives an AI startup a moat, or what makes it different compared to its peers. More than half of the respondents said that the thing that gives AI startups an edge is the quality or rarity of their proprietary data.

Paul Drews, a managing partner at Salesforce Ventures, told TechCrunch that it is very difficult for AI startups to have a moat because the landscape is changing so quickly. He added that he looks for startups with a combination of differentiated data, technical research innovation, and a compelling user experience.

Jason Mendel, a venture investor at Battery Ventures, agrees that technology moats are shrinking. “I look for companies with deep data and workflow moats,” Mendel told TechCrunch. “Access to unique, proprietary data enables companies to deliver better products than their competitors, while a sticky workflow or user experience allows them to become core systems of engagement and intelligence. that customers rely on every day.”

Having proprietary, or hard-to-get, data is becoming more important for companies building vertical solutions. Scott Beechuk, a partner at Norwest Venture Partners, said the companies that can house their unique data are the startups with the most long-term potential.

Andrew Ferguson, vice president of Databricks Ventures, says that having a lot of customer data, and the data that creates a feedback loop in an AI system, makes it more effective and helps startups stand out as well.

Valeria Kogan, the CEO of pausea startup that uses computer vision to detect pests and diseases in plants, told TechCrunch that he thinks one of the reasons Fermata is gaining traction is that its model is trained from customer data and data from the company’s own research and development. center. The fact that the company does all of its data labeling in-house also helps make a difference when it comes to model accuracy, Kogan added.

Jonathan Lehr, a co-founder and general partner at Work-Bench, added that it’s not just the data that companies have but how they can clean it and use it. “As a pureplay seed fund, we focus most of our energy on vertical AI opportunities that address business-specific workflows that require deep domain expertise and where the AI is basically a factor in acquiring previously inaccessible (or very expensive to obtain) data and cleaning it in a way that would take hundreds or thousands of human hours,” Lehr said.

Beyond data, VCs say they’re looking for AI teams led by strong talent, those with existing strong integrations with other technologies, and companies with a deep understanding of the flows. of the customer’s work.