Founded in 2014, Behavox is a global leader in AI-powered compliance, security, and archiving solutions. Headquartered in London with offices in New York, Montreal, Singapore, the company helps organisations secure their communications data and monitor regulatory, conduct, and insider threat risks.
As a pioneer in RegTech, Behavox has spent the last decade growing from strength to strength. The past year, in particular, has marked a pivotal moment for both the company and the broader industry.
According to Nabeel Ebrahim, Chief Revenue Officer at Behavox, 2024 is shaping up to be a turning point. “We’ve seen a surge in regulatory interest around AI,” Ebrahim explains. “Regulators like the CFTC are seeking input on AI, and organisations like FINRA are endorsing its experimentation. The industry is catching up to the technology we’ve built, and we’re in a prime position to lead.”
Solving Critical Pain Points
In a multifaceted industry like RegTech, addressing key pain points is essential for success. Behavox has focused on solving three major challenges facing the financial sector.
The first challenge is the growing risk profile in the industry. “As organisations grow, so do the risks,” Ebrahim notes. “The number of employees, languages they use, and communication channels they rely on—like Microsoft Teams and other messaging platforms—have expanded. During the COVID-19 pandemic, we saw a rise in these channels, leading to $2.8 billion in fines for improper use.”
Behavox solves this by natively integrating and ingesting communication channels while addressing multiple languages—49 to be exact—through proprietary large language models specifically built for financial services.
The second challenge is reducing operational overhead, particularly around false positives. “Organisations are drowning in false positives,” Ebrahim says. “Legacy technology should’ve evolved past this point, but it hasn’t. Our solution reduces the number of false positives and minimises the hours and manpower needed to manage them.”
The third pain point focuses on risk identification. “It’s not enough to reduce false positives,” Ebrahim explains. “People have become smarter in how they communicate. You need a financial services-specific AI model that can identify and capture true risks, and that’s what we’ve built.”
Prioritising AI Explainability
As AI becomes more ingrained in various industries, the need for explainability grows. Companies must not only leverage AI but also ensure regulators and stakeholders understand how it works.
“Explainability is a key requirement for any model risk management regulation,” says Ebrahim. “In 2011, the Federal Reserve set model validation guidelines, and now we have similar regulations coming from the Bank of England and the EU’s AI Act.”
At Behavox, explainability is central to its approach. The company spends significant time providing clients with straightforward explanations and technical insights into how its AI models are built, tested, and documented. Customers can independently test and validate the AI to ensure it meets their needs.
Ebrahim adds, “This process, known as outcomes analysis, allows organisations to understand not only how the AI functions but how it works for them specifically. It’s a key part of internal audits, external audits, and regulatory compliance.”
Building Industry-Specific AI Models
A major point of differentiation for Behavox is its focus on developing domain-specific AI models. “Generalist AI models aren’t designed to detect financial crimes like momentum ignition or collusion,” Ebrahim explains. “These models might help with writing emails or organising a workout, but they aren’t built for banking.”
Behavox has taken a different approach by developing large language models that are fine-tuned to detect specific risks relevant to financial services. “We’re building AI that acts like a highly specialised financial analyst, able to identify the nuances of risk and catch the ‘bad apples’ within organisations,” says Ebrahim.
Leveraging AI and Overcoming Challenges
As AI becomes ubiquitous across industries, companies are learning the best practices for implementation and use. According to Ebrahim, enterprises must take a structured approach to both.
“There are two sides to best practices,” he says. “First, when implementing AI, organisations need to experiment and benchmark various approaches. Don’t just take a vendor’s word for it—ask for explainability and evidence of how it compares to what you currently use.”
Second, Ebrahim stresses the importance of continuous improvement. “It’s vital to establish a feedback loop, where users monitor the AI’s results and work with the vendor to refine and improve the system. AI is not a ‘set-it-and-forget-it’ solution; it evolves, and companies need to evolve with it.”
When it comes to challenges, Ebrahim points to scepticism as a major hurdle. “Rational sceptics are often cautious about adopting new technology, especially if they’ve been using the same methods for 15 or 20 years. The challenge is that by the time they decide to move forward, competitors may have already pulled ahead.”
Another significant challenge is data governance. “AI is only as good as the data it works with,” Ebrahim says. “If your data is poorly managed, even the best AI won’t perform well. Before applying AI, organisations need to clean and understand their data.”
Key Trends in AI
As AI continues to evolve, certain trends are emerging, particularly around language models. “It’s not about bigger models anymore,” Ebrahim observes. “While large generalist models from companies like OpenAI and Google are impressive, we’re seeing a shift towards miniaturisation and specificity.”
According to Ebrahim, organisations are increasingly building their own models or partnering with companies that provide highly specialised AI tailored to specific industries or challenges. “The future lies in creating AI models that solve precise problems with high accuracy.”
Another trend is the growing scrutiny around explainability and testing. “Many companies are trying to repackage general AI solutions like Copilot under different names, but these will fall out of favour as model validation teams demand more transparency and accountability,” Ebrahim predicts.
The Road Ahead for Behavox
Looking ahead, Behavox plans to continue expanding its AI offerings, including a new initiative called Behavox Pathfinder. “We’ve built our own version of ChatGPT, tailored specifically for financial institutions,” Ebrahim explains. “This allows clients to upload their policies and research into a dedicated chatbot, making it easier to access critical information quickly.”
With Behavox Pathfinder, the company aims to increase efficiency and reduce the time employees spend searching for information. “Whether it’s finding an expense policy or referencing internal research, we’re working to streamline these processes and deliver real value.”
As Behavox continues to push the boundaries of AI in RegTech, the company is poised to remain at the forefront of compliance and security innovation.
Copyright © 2024 RegTech Analyst
Copyright © 2018 RegTech Analyst
