• Financial companies and fintech startups are trying to use generative AI to improve search capabilities.
• Search is a complex and difficult technical problem, but AI could make it much easier.
• Here’s how companies like Goldman Sachs and Blackstone are trying to crack the research code.

If data is the new oil, then banks like Goldman Sachs are in good shape.

For decades, Goldman collected information about the clients it did business with, the trades executed, every dollar invested, and every loan funded. Coupled with external data from Bloomberg and Nasdaq, the hope was to supercharge the bank’s analytics engine and give an edge to its investment bankers, traders and salespeople.

But this fuel is only useful if you can access it.

For banks, much of this data was stored, usually only to be found if an employee knew exactly what they were looking for and where to find it.

All that could change as Goldman Sachs rolls out an AI generative chat interface to its company-wide data platform, Neema Raphael, chief data officer, told Business Insider.

Goldman employees can ask a question in plain English and let the AI ​​dig. By answering questions, the tool, called Legend AI Query, could extract information that even its users didn’t know existed.

The chat interface, combined with the bank’s data stores, “gives you this type of super intelligent information to help the human build a better mental model faster and faster with more sources,” Raphael said .

It’s the latest development in Wall Street’s efforts to crack the research code.

From Goldman Sachs to Blackstone, the biggest financial companies are using generative AI to make better use of their mountains of data. Although it has been decades since Google introduced the world to effective search for use in everyday life, it is only recently that financial companies have begun to devote resources to improving the way that employees use their internal data. They attempt to turn the tricky and sometimes impossible task of searching for information into a seamless process that will boost employee productivity. Research refinement, over time, could lead to more automation and more complex generative AI tools.

Research is just the beginning

Goldman’s peers across the street have their own research initiatives underway. JPMorgan’s Private Banking AI Co-Pilot helps advisors find insights in real time. Bank of America’s Banker Assist aggregates first-party and third-party information to give employees insights. Morgan Stanley’s AIMS helps advisors search the bank’s internal content.

Although enabling employees to quickly get answers hidden in mountains of data will increase worker productivity, there is likely even greater ambition behind these efforts.

Extracting the right information and having some understanding of context is the first step to exploiting more complex use cases, Keri Smith told BI. Smith helps financial companies strategize and execute their data and generative AI initiatives at Accenture.

“The power of enterprise search is its ability to save time so humans can innovate and interact,” Jeff McMillan, head of enterprise-wide AI at Morgan Stanley, told BI . “Additionally, it reduces the barriers that prevent employees from quickly accessing solid intellectual capital from the company’s top experts, essentially equipping them with knowledge firepower for meetings and discussions,” he added.

A new class of fintechs is beginning to emerge to sell Wall Street by removing simple but time-consuming tasks, like perfecting corporate logos at an investment bank and grooming executives before client meetings.

Rogo is one such startup that offers a junior banker level assistant. It has already onboarded some 25 Wall Street firms on its generative AI platform.

“Companies are realizing the value of enterprise search not just for a traditional search engine, but also for all the downstream applications you can build off of it,” said Gabe Stengel, CEO of Enterprise Search. co-founder and CEO of Rogo, at BI.

Meanwhile, two Stanford graduates came together to create Mako, a generative AI associate for the private equity industry. The startup, which aims to help employees search for institutional data, recently raised $1.55 million from the same venture capitalist that was an early backer of OpenAI.

Why research is difficult

There’s a reason why Google is the go-to search engine on the Internet.

“It’s actually a fundamentally difficult problem to sort out and then sort out what might be useful” to a specific user, Raphael said of the research. Not only is search a difficult IT problem, but customizing relevance and managing account permissions (who is allowed to see the given data) is not an easy task, he added.

The latter is something Blackstone discovered for the majority of 10 months when it recently built its own internal AI-powered search engine.

Add to that the fact that Wall Street jargon is complex and nuanced – words like hedge, ticker and options have completely different meanings outside of a financial context – and it presents another obstacle for financial firms using products commercially available, such as OpenAI’s ChatGPT.

In March, Balyasny Asset Management hired Peter Anderson, a former AI scientist at Google, to help the hedge fund improve its back-end system that extracts information from millions of documents to answer complex research questions. Familiarizing OpenAI’s models with financial jargon allowed Balyasny’s internal version of ChatGPT to surface the most useful document 60% more frequently than without this training.

Yet generative AI brings companies closer to solving search problems.

“People are building knowledge bases, they’re letting this genAI explore and being able to do searches or summaries, I think maybe that’s a stepping stone to solving the problem,” Raphael said.

Where Goldman is putting its weight behind generative AI

Legend AI Query is just the beginning for Goldman Sachs.

The search tool is the bank’s second generative AI tool, the first being a generative AI developer co-pilot that helps software engineers code more efficiently. This effort resulted in approximately a 20% increase in efficiency depending on the use case.

As Goldman’s AI/ML engineers worked to crack the research code, they also created another generative AI tool that aims to help data engineers – the developers who manage the bank’s data and make sure they are vetted, organized, and structured — do their jobs better. Legend Copilot is another tool launched by the bank this month, designed to make it easier to get more data about Legend and manage it in a methodical way.

Raphael said he’s focused on “really helping engineers and non-engineers alike find and discover the right data for their use case.”