See the 14-page pitch deck that sold Citi on Claira, a startup using AI to help firms read through financial contracts in a fraction of the time
This story is available exclusively to Business Insider subscribers. Become an Insider and start reading now.
- Claira is a startup that provides AI for financial contract analysis.
- It was born out of Exos, an investment bank founded by a former Citadel CTO and Credit Suisse CEO.
- The upstart scored Citibank as a client and funding from the bank’s strategic investment arm.
It was a match made in heaven — at least the Wall Street type.
Joseph Squeri, a former CIO at Citadel and Barclays, had always struggled with the digitization of financial documents. When he was tapped by Brady Dougan, the former chief executive of Credit Suisse, to build out an all-digital investment bank in Exos, Squeri spent the first year getting let down by more than a dozen tools that lacked a depth in financial legal documents.
His solution came in the form of Alex Schumacher and Eric Chang who had the tech and financial expertise, respectively, to build the tool he needed.
Schumacher is an expert in natural-language processing and natural-language understanding, having specialized in turning unstructured text into useful business information.
Chang spent a decade as a trader and investment strategist at Goldman Sachs, BlackRock, and AQR. He developed a familiarity with the kinds of financial documents Squeri wanted to digitize, such as the terms and conditions information from SEC filings and publicly traded securities and transactions, like municipal bonds and collateralized loan obligations (CLOs).
The three converged at Exos, Squeri as its COO and CTO, Schumacher as the lead data scientist, and Chang as head of tech and strategy.
Squeri, who was working alongside Chang in finding a vendor partner, asked Schumacher why there was nothing in the market that provided deep insights, transparency, and accuracy in a timely manner without training.
“He says, ‘Well it’s because they’re all using the same open source libraries and the same approaches and no one has rethought the problem,'” Squeri told Insider of Schumacher’s response. “Eric and I looked at each other and said, ‘Oh, here lies the opportunity,'” he added.
Two weeks later, Schumacher had a prototype and Claira was born.
And opportunity it was — Exos went on to save 90% of time and effort on the investment bank’s LIBOR remediation and has expanded the tech to other use cases.
Meanwhile, Citibank has since signed on to use the tech and make a strategic investment in the startup.
A new approach wins over Citi
Claira’s underlying framework uses a field of data science called computational linguistics that essentially breaks down documents into sentences. The sentences, entities, and relationships between the entities are translated into a map that machines can work with, Chang said.
Then, depending on who’s using the tech — from a municipal bond trader to an investor — analytics are customized and pertinent info is pulled and summarized.
“Traditional approaches to the digitization of documents has been the natural-language processing, machine-learning approach of data science, which really only gets you so far,” Squeri said.
That’s because most systems work by pattern matching, or turning sentences and paragraphs into numbers. But there are shortcomings to the approach. When AI provides an answer in a black box there is no way to trace back how the answer was derived, Squeri said. Legal language is also sometimes ambiguous, meaning it doesn’t translate to a one or a zero, he added.
Because Claira starts at the sentence and builds a map from there, each insight can be traced back to the exact sentences used, Chang said.
For example, the bot can ingest a 500-page document and instead of an analyst using keyboard commands to find certain terms or reading the whole package, Claira will essentially highlight certain sentences that the user needs to read.
Exos was the first finance firm to use the product for its remediation of the LIBOR index at the end of 2020. The company saved 90% of time and effort compared with using outside counsel to read the documents and send back an Excel spreadsheet, Squeri said. And the accuracy was higher with the AI compared to using humans, whose accuracy rates for reading documents and annotating agreements range from 85% to 90%, Squeri added.
Since then, Citibank has signed on to use the tech in its capital markets group to provide a lift to the trading desks dealing with CLOs and municipal bonds, Chang said. The bank’s traders use Claira to get data around provisions found in the offering memorandum for CLOs and offering statements for municipal bonds to better price and measure risk for each transaction.
As part of the deal, Citi’s strategic investment arm, SPRINT (spread products investment technology), invested an undisclosed sum in Claira. Chang declined to disclose the startup’s total funding.