AI Firms Are Racing to Take Over Wall Street's Late-Night Grunt Work

Rogo, a startup that builds AI tools for banks, has passed $50 million in yearly recurring revenue, up from $15 million at the end of 2025, according to a PYMNTS report citing The Information. OpenAI and Anthropic are going after the same work: the financial models, memos and pitch decks that junior bankers often build late at night.

What changed

AI is moving past answering questions. It is now preparing financial models, analyzing deals and producing investment research. Anthropic debuted 10 finance agents in May. An agent is AI software that carries out a task instead of just chatting. OpenAI followed in September with ChatGPT for Financial Services, built with Morgan Stanley and Evercore as design partners.

Three kinds of players

• Data owners such as LSEG, Daloopa and PitchBook supply the raw information.
• Workflow startups such as Rogo and Hebbia turn it into finished work.
• Content owners such as Bloomberg and AlphaSense build their own AI tools.

The data owners sell to both the big AI labs and the startups. That means they get paid whoever wins. PitchBook, which Morningstar owns, put its private market data into ChatGPT in December. It had already connected to Claude, Perplexity and Microsoft's Copilot Studio.

Where the money is going

Banks are buying. Rogo's clients include Lazard, J.P.Morgan, Bank of America and Wells Fargo. Hebbia is approaching $50 million in annual recurring revenue and holds a contract covering about 5,000 Morgan Stanley investment banking employees. AlphaSense raised $350 million in June at a $7.5 billion valuation.

The biggest check so far went to data. On Oct. 22, LSEG agreed to buy credit data provider Octus for $3.5 billion. Bloomberg also made ASKB, its conversational AI tool, generally available on Oct. 15. It runs partly on Anthropic models.

What is still unclear

Rogo's CEO, Gabe Stengel, said a lot of Wall Street analysis is "done by a 21-year-old in tools from 40 years ago at 2 a.m." His software aims to take that off the analyst. The material does not say how many entry-level jobs could change. It also leaves open who will capture most of the value: the labs, the startups or the data owners.

Farsight CEO Samir Dutta argues banks still lack clear enough use cases to build good tools in-house. That is the startups' case for staying in business against the labs.