Regulatory compliance is becoming a costly challenge for financial institutions: A recent survey found that 76% of financial services companies are increasing their compliance spending from 2022 to 2023, mostly due to new legislation.
With compliance costs averaging roughly $10,000 per employee these days, many companies are looking for ways to cut spending without running afoul of regulators. Entrepreneurs Nir Laznik and Eyal Peleg say they've followed the trend and developed a solution that leverages generative AI.
Lasnick and Peleg co-founded Sedric, an AI-powered platform designed to help financial institutions enforce compliance rules and flag potential issues. Prior to founding Sedric, Lasnick led several startups, including a photo kiosk software company, and Peleg worked in Intel's AI and machine learning organization for about eight years.
“We realized that mid-sized organizations were under disproportionate pressure and that banks had a number of new challenges,” Lasnick says, “and we knew that the rapid advances in AI could help address these issues in entirely new ways. It was from this confluence of factors that Sedric was born.”
Sedric's AI acts as a sort of supervisor, monitoring employees' inbound and outbound calls, chats, emails, social media DMs, and instant messages. It flags compliance issues as soon as they arise (missed disclosures, omitted procedures, fraud, etc.). Laznik claims that Sedric can often “mitigate” issues automatically, offering guidance to non-compliant staff.
“The technology gives compliance officers a holistic view of customer touchpoints across multiple channels, enabling them to quickly and efficiently detect deviations from established compliance policies and guidelines,” Lasnick said. “Our platform covers the entire compliance lifecycle, from policy setting to enforcement, remediation and auditing.”
Screenshot of Sedric's backend dashboard. Image courtesy of Sedric
Such extensive monitoring might seem a bit intrusive, since Cedric “scores” each employee's interactions according to their compliance with company policy, but for better or worse, US state and federal guidelines give companies wide leeway in engaging in employee monitoring, so long as they're fully transparent about it.
Additionally, federal regulations, particularly those regarding insider trading, collusion, and the sharing of certain earnings documents, require financial institutions to closely track their employees' interactions with clients and the market as a whole. These supersede state laws such as those in New York and Connecticut that impose additional requirements on employers who monitor their employees.
Because Cedric's AI will likely monitor communications between staff members from a variety of backgrounds, I asked Lasnick about the possibility of bias in the company's AI. Biased AI, whether intentional or not, could lead to discrimination depending on where and how it's deployed.
Studies have found that some AI trained to detect harm tend to disproportionately label English phrases in African American slang, an informal grammar used by some Black Americans, as “harmful.” Other studies have demonstrated that speech recognition systems are more likely to mistranscribe the audio of Black speakers than White speakers.
Cedric tries to minimize bias by using “fine-tuned models” trained on “proprietary datasets that are curated and validated in collaboration with industry experts,” Lasnick said. The company also monitors deployed models for performance degradation and retrains them when necessary.
Image credit: Cedric
“Our platform allows customers to provide direct feedback through a variety of annotation inputs, which is vetted by our compliance team and used for retraining or integrated into the prediction process,” he added. “This allows our models to become increasingly customized for each customer.”
To protect the privacy and security of its customers and employees, Sedric allows companies to implement controls to configure where data is stored and to redact (or at least attempt to redact) personally identifiable information.
“At Sedric, we designed our platform with compliance and security at its core,” says Lazink. “Companies can set their own retention policies and compliance guidelines in accordance with their internal guidelines and specific regulations.”
Cedric also provides tools to support call centre agents when chatting with customers over the phone, and Lasnick claims to have “hundreds” of paid compliance officers and corporate clients across the US and Europe.
Revenue is up five-fold from last year, though Lasnick declined to provide more precise figures.
“We offer off-the-shelf solutions for SMEs and customized hybrid models for larger enterprises and banks,” Lazink said. “We stand out in the market by focusing on the specific needs of financial institutions, combined with a unique library of pre-trained, regulatory-inspired models that can be customized to fit each organization's unique requirements.”
Targeting financial customers and use cases specifically has certainly worked to Cedric's advantage, setting it apart from workplace monitoring rivals such as Fairwards, Shield, Erudit and Aware. While the market is competitive and often contentious, investors still see some opportunity, especially as AI becomes more deeply embedded in these tools.
For one, Foundation Capital, seemingly pleased with Cedric's progress so far, led an $18.5 million Series A investment in the four-year-old company, with participation from Amex Ventures. Radzink said the new funding will be used to “significantly” expand the company's go-to-market and research and development teams in New York City and Tel Aviv, bringing New York-based Cedric's total funding to $22 million.
Cedric plans to double its workforce within the next 12 months.