Frontier Al labs are slowly encroaching on well-funded startups that have served companies’ industry-specific AI needs. But making inroads is slowgoing and some corporate clients are locked in by existing products.

Take investment banking, for example. OpenAl hired more than 100 former investment bankers to train its financial products last year, and Anthropic this week launched agents for reviewing balance sheets and drafting pitch decks. Similar products are already live in investment banks via companies like Rogo, the buzzy $2 billion startup automating much of the junior analysts’ and associates’ workload at firms like Tiger Global, Jefferies, and Lazard.

One banker at a firm who uses Rogo told Semafor even if OpenAl released a better product, Rogo is already integrated in employees’ workflows. Rogo, which operates with models from OpenAI and Anthropic, among others, has dozens of bankers working with its customers. For companies like OpenAl and Anthropic, competing in that limited space is human- and capital-intensive, and it means playing catch-up.

“There’s so much to build outside of the core model intelligence,” Rogo CEO Gabe Stengel told Semafor recently, when asked about competing more closely with the frontier labs. “That’s not the core competency of these labs who, frankly, don’t understand investment banking.”

More competition is always good though, yielding quicker product releases and better customer service.