Generative-AI startups are receiving billions of dollars in funding, but a critical challenge could derail their path to success. These firms require the right data to develop the algorithms and models necessary to build competitive business applications. However, obtaining proprietary data is no easy feat. Brad Svrluga, co-founder and general partner of Primary Venture Partners, noted that there are plenty of startups with brilliant AI applications but no access to data. Paul Tyma, CTO-in-residence of Bullpen Capital echoed these thoughts, saying that having the right data is now more critical than ever before. He added that building models is somewhat commoditised right now, but the value lies in the data.
According to PitchBook, venture funding in generative-AI has grown from $4.8bn in 2022 to $12.7bn in the first five months of 2023. Many of these startups also face hurdles when searching for training data sets in finance and healthcare.
Furthermore, AI startups are seeking partnerships with big, data-rich companies to overcome this problem. They are targeting firms such as EY, with its vast transactional data, to solve their data woes. However, Andy Baldwin, Managing Partner of Client Service at EY, voiced concerns about what would happen to the company’s data if it were used to train an external model.
To sidestep these IP concerns, startups have begun training separate models for their clients, using only client data. This strategy allows ventures such as TermSheet to build its generative-AI model Ethan, which answers industry questions posed by real-estate developers, brokers, and investors. However, it is contingent on convincing customers to agree to this arrangement.
The difficulties of ensuring cybersecurity protection further add to these challenges. Obtainment of proprietary data becomes even harder for early-stage startups without the recognition and social proof larger firms enjoy.
This situation favours large tech incumbents with strong track records. Tracy Daniels, Chief Data Officer of Truist, expressed her preference for exploring generative AI use cases with larger tech vendors rather than startups, citing trust and data protection concerns.
In conclusion, generative-AI startups have a critical problem on their hands; without proper access to data, they stand to face failure. While partnerships with data-rich incumbents can help ameliorate some of these issues, startups also need to convince their clients about the safety and confidentiality of their data while ramping up cybersecurity measures. Struck Capital’s founder, Adam Struck, also sees startups competing to secure more data faster, making data a critical differentiator in the AI space.
