How is financial services leveraging AI for compliance purposes?

AI

In a time where AI is revolutionising industry after industry, the increasing need and desire to automate compliance is becoming ever more sought after.

The role of AI in financial services is rapidly increasing, with the introduction of ChatGPT causing deep ripples in the financial industry. With compliance being a permanent challenge for companies to maintain, the ability to automate this challenge provides significant upsides for businesses.

“We’re seeing AI play a fundamental role in enabling financial services companies to supervise their business communications,” said Stacey English, director of regulatory intelligence at Theta Lake. She highlighted that with the modern workplace now powered by modern unified collaborations like Zoom and Teams, many regulated organisations are facing a huge challenge in capturing and supervising these complex multimodal capabilities.

Theta Lake provides compliance and security for modern collaboration platforms and can capture, compliantly archive, and act as an archive connector for existing archives of record for video, voice, and chat collaboration systems. “Our financial services customers are leveraging AI to review vast volumes of communications which would otherwise outstrip the capacity of their compliance teams to capture and manually review them,” explained English.

The company’s technology also captures and analyses all of the key contextual and rich media such as GIFs, edits and deletes, which can change the meaning of conversations. English stated, “The AI not only identifies which communications contain regulatory, privacy or security risks but pinpoints exactly where the potential breach occurred saving reviewers having to trawl through entire meetings or chat conversations.”

English noted that the industry has seen a real shift from legacy approaches to supervision  which was designed for the ‘now ancient email era’. She said that prior approaches to supervision relied on ‘highly manual and technically simplistic word-searches and lexicons’, while modern approaches to AI-enabled supervision facilitates more sophisticated analysis of data to identify issues by analysing content in context, to understand where and when financial services-relevant conversations are taking place.

How does English see this space evolving? “AI can only be effective in supervision and risk detection if it has comprehensive data to analyse, so the complete capture of communications without any blind spots is integral to leveraging AI”. Going forward, she explains, the industry can expect even greater demand for assurance that they are capturing all channels of communications, alongside all of the rich contextual information.

“Furthermore, there will no doubt be increasing requirements for visibility into AI as Boards and regulators seek assurance. At Theta Lake, we’ve taken several steps to ensure that customers have transparency, including audit reports that demonstrate review processes, indicate where the AI has been triggered and allows firms to oversee AI’s performance.

“Ultimately, as a vendor it’s really important to be involved in industry and regulatory dialogue to inform the future state of AI, so we were pleased to provide feedback to the FCA and PRA’s latest consultation on the use of artificial intelligence and machine learning in financial services.”

Knowing its use

One of the key challenges for many around the use of AI is how best to exploit it for the needs of a company – the rise of platforms such as ChatGPT are a sharp example of this.  In the opinion of Aveni COO Jamie Hunter, the recent surge of these platforms are ‘raising the stakes’ both for the tech and how businesses chose to use it.

He commented, “In financial services the use of AI is not new, especially in larger organisations, but it has been pretty limited to fairly low risk scenarios and not fully embedded across business models until this point. That is changing as large language models and natural language processing evolve and advance, and this will be expedited in financial services as the emphasis on Consumer Duty and proving customer outcomes through a data-first approach grows.”

Hunter explained how risk management and compliance have ‘traditionally been seen as an afterthought or an inconvenient necessity’ rather than a value-filled function. With new regulatory demands, technology is enabling the compliance function to sit as the central nervous system of an organisation by using AI to mine data that can drive business decisions in multiple functional areas.

“The use of natural language processing with AI is squarely putting the voice of the customer first – the outputs are real and specific to what the customers are saying. This means that complaints, concerns and requests from customers as well as the advice they are being offered, can be monitored fully and assessed accurately,” he explained.

Hunter believes that there are many key areas where AI is starting to change compliance – however, this continues to evolve and will depend not only on the investment made but the priority it is given and the importance of identifying the right data and inputs to ensure the most effective outputs.

He continued, “It will guarantee better identification of vulnerable customers and allows businesses to be proactive rather than reactive, identifying trends using data before they become problems. It is already, and will continue to enhance management information and reporting, and allow far greater coverage of customer service calls and interactions.

“It is generating efficiencies through automated Quality Assurance and also can help identify where more training and development is required for staff interacting with customers. Technology is also enabling a greater defence in regulatory compliance, capturing every single customer interaction and analysing it to provide vital feedback.”

Hunter concluded by stating that he believes large language models will be able to outperform humans in every economically-valuable task within a financial services organisation in the next ten years – and the industry needs to ensure the adoption of this is done properly with ‘greater collaboration between the engineers and the financial experts to make this as effective as it can possibly be’.

A key use of AI in financial services for a while has been its use to support decision making in a wide range of risk scenarios. However, James Brodhurst, principal consultant at Resistant AI, believes financial crime compliance is one area where it could be said that adoption has been slower than for other financial crime use cases.

He stated that the reasons are several, but include some of the broader misconceptions about AI, such as that it always requires a huge historical sample of data to be trained, or that the results may not be easily explained to a regulator and even disagreement on what constitutes a good result.

“Despite this, AI does already have a role in improving compliance and we expect this to evolve significantly in the near future,” said Brodhurst. “Adopting AI means assessing carefully the goals of such an initiative. Whilst some FinTechs have made the leap from more rules-driven compliance defences to fully AI-powered detection we believe it will be more common for some time yet to see AI augmenting existing platforms, working alongside existing due diligence capabilities and filling the voids that powerful analytics is best placed to cover.”

Rather than taking over completely, Brodhurst claims that he sees AI being used to allow compliance teams to focus on truly value-added work.

He concluded, “At a time when emerging technologies such as generative AI (such as ChatGPT) offer powerful support to financial criminals, the ability to keep track with evolving crime patterns has never been more urgent. AI is able to rise to this challenge with techniques that allow FinTechs to uncover novel patterns as they develop.”

Machine learning revolution

The rise of machine learning over recent years has seen its presence in company’s technological offerings increase endlessly. As Vladimir Ershov – head of data sciences and machine learning at Clausematch – states, its adoption in financial services has been driven mainly by use cases related to fraud and AML.

He continued, “However, the automating of tasks like gap analysis and contradiction detection in the entire corpus of a company’s policies has long been considered an unreachable dream because of the fundamental challenges of natural language processing tasks.

“But today, we’re witnessing mind-blowing progress that has changed the game overnight. In 2018, the introduction of BERT and GPT-2 signalled that we finally had technologies capable of performing deep near-human analysis of texts. Subsequent breakthroughs, including ChatGPT and GPT-4, have showcased not only near-human intellect, but also the reasonable cost of execution and a context window large enough to analyse a complete document.”

Ershov remarked that these breakthroughs, packaged today into infrastructure, represent a ‘complete revolution’ in document processing. “As we see with the unveiling of GPT-4 earlier this month, there’s no compliance task that can’t be automated, up to at least 95% in principle. evolution of AI in the financial services compliance space is exponential, and the slope of progress is already very steep,” he expressed.

Regulatory compliance

 One of the most critical aspects of the financial services industry is regulatory compliance, with many complex regulations and increased regulatory requirements for financial firms meaning that a lot of resources and expertise are used to ensure compliance with a changing regulatory landscape.

According to MAP S.Platis, the integration of AI in financial services compliance, can significantly transform traditional processes and help financial firms to simplify their processes and enhance their compliance practices.

The firm said, “Key AI-powered tools for financial services compliance which are currently in use by Financial Firms, such as the KYC, AML transaction and monitoring systems and market surveillance, enable financial firms to speed up their transaction screening/review processes to identify suspicious transactions and potential risks in real-time, eliminating the false alerts and mitigating risks while saving time and resources.”

The application of AI in compliance is also able to boost the efficiency of compliance operations which traditionally relied on manual processes – many of these tools powered by AI can help compliance officers with real-time alerts and notifications, enabling them to react swiftly to potential compliance violations.

MAP S.Platis added, “Nevertheless, given the strength and capacity of the human brain in terms of learning, comprehension, analysis, rationalism, and decision-making, the human factor remains critical in compliance space and (for now) cannot be completely replaced by AI-powered tools.

“Also, firms remain accountable for the final decision, which cannot be attributed to AI systems, thus the human intelligence remains critical but can heavily be supported by the adoption and integration of AI-powered technology.”

MAP S.Platis explained that it believes AI is playing a ‘critical role’ in the field of financial services compliance, and its significance is only expected to climb as regulations become more complex and the volume of data continues to increase.

“As the technology evolves, more advanced AI-powered compliance solutions can be expected, which will provide even greater efficiency, accuracy, and insights into compliance-related activities, with the human factor remaining critical and accountable in the final decision-making process,” the firm concluded.

Compliance efficiency

In a world where the challenges faced by financial services businesses are multiplying rapidly, the need for efficiency in compliance becomes an ever more welcoming desire.

According to Joe Schifano – global head of regulatory affairs at Eventus – financial services has been leveraging AI for compliance purposes in various ways. “One of the most significant roles that AI is playing in this space is automating and streamlining the compliance process. AI can help financial institutions sift through vast amounts of data to identify potential compliance violations, reducing the risk of human error and increasing efficiency,” he claims.

In addition, he stated that another key role AI is playing in the financial services compliance space is fraud detection. He stated, “AI algorithms can identify unusual patterns in transactions and alert compliance officers to potential fraudulent activities, which can help prevent financial crimes.”

He continued, “The use of AI in financial services compliance has changed significantly in recent years. It has moved beyond traditional rule-based systems to more sophisticated machine learning models that can analyse vast amounts of data in real-time. This has enabled financial institutions to stay ahead of compliance issues and reduce the risk of regulatory violations.”

“In the future, we can expect to see even more advanced AI-powered compliance tools that can adapt and learn from new data, as well as natural language processing capabilities that can analyze unstructured data such as emails and social media. As regulations continue to evolve, financial institutions will need to continue to leverage AI to ensure they remain compliant while also reducing the cost and time involved in the compliance process.”

Managing complexity

The increasing number of regulations in the financial industry mean that any technology that is introduced is focusing on increasing efficiency and reducing complexity. Joseph Ibitola, growth manager at Flagright, believes that both AI and ML techniques hold great promise in helping financial institutions find efficiencies in this area.

He explained, “Their goal is to deploy AI/ML algorithms to improve automation, speed and accuracy. Regulatory and compliance staff can leverage AI/ML algorithms to address risks quicker and focus on the most important issues that threaten their firm’s regulatory standing and reputation.”

Ibitola cited ‘promising’ use cases in production today such as AI/ML algorithms that compute probability scoring of alerts, algorithms that detect abnormal trading patterns through anomaly detection and NLP to monitor messaging platforms and media channels for suspicious activity.

He continued, “Compliance teams need to be mindful of both what AI/ML algorithms do and how to deploy them as part of a wider set of technologies. Regulators globally require compliance officers to “explain” how AI/ML models determine which activities to promote as an alert.

“Early use cases are promising and financial institutions will continue exploring use cases and deploying the latest technology to improve efficiency. This is becoming increasingly critical as regulators globally are doing the same—financial institutions will want to keep pace.”

Consistency and structure

As a sector, financial services has always been a heavily regulated space – with compliance requirements consistently changing. In order to comply with these regulations, many financial institutions have traditionally relied on manual processes to review vast amounts of documents and data – something that makes the process time-consuming and costly. However, Jon Leitner – president of Ascent – believes the rise of AI has made compliance processes become more efficient, accurate and cost-effective.

He continued, “One significant way AI is transforming the financial services compliance space is through its ability to organize vast sets of information and documents. AI-powered tools can ingest large volumes of data, categorize, tag, and index it, making it easier to locate and analyse. This capability can be particularly useful for regulatory compliance, as it allows financial institutions to quickly identify and track specific transactions, customers, or patterns of behaviour.”

Leitner also suggested that AI is able to bring consistency to the data structure itself, which can therefore remove subjectively from people’s points of view. This standardisation, he claims, can be crucial in ensuring data is consistent across different departments and can be compared easily. This can substantially reduce the potential for errors or omissions in compliance reports.

“AI can provide financial institutions with greater flexibility in adapting to changing compliance requirements. AI-powered compliance tools can be trained on new regulations and requirements, and they can quickly incorporate updates and changes as they occur. This can be particularly valuable in an environment where regulations are constantly evolving and can reduce the risk of non-compliance penalties,” Leitner continued.

He concluded that the technology is ‘playing an increasingly important role’ in the financial services compliance space by helping institutions to better organise vast sets of information and documents. “As AI technology continues to evolve, we can expect to see even greater improvements in compliance efficiency and accuracy.”

Muinmos CEO Remonda Kirketerp-Møller also commented, “AI technologies have significantly improved compliance, by improving both the quality of the examinations made and their quantity. Simply put, machines can do certain things, like ID verification, better than humans. They can assess things we cannot, like MRZ (Machine Readable Zone) and various other factors the human eye is not trained to see. They don’t get tired and make errors. And they do it in the blink of an eye, allowing each identity to be verified.

AI has also allowed for better monitoring, for example of transactions, and subsequently detection of potential violations in real-time. By continuously analysing large amounts of data, these systems can identify patterns and anomalies that might indicate a compliance issue. AI solutions also exist in Risk Management, regulatory reporting and more. In each field, AI has made compliance more accurate, fast and complete.”

Kirketerp-Møller remarked that when the company incorporated AI into its product a decade ago, ‘it was innovative’. Today, however, she claims that a company cannot do compliance without AI – it is the base-standard for compliance. 

“And now, with the new language-models like Chat GPT4, this reliance on AI will only increase. The danger is, of course, that we give AI full control of compliance, and forego human oversight. This danger is being addressed now by regulators worldwide, ensuring the AIs used are non-biased and transparent (meaning, that the logic behind their decisions is known).

“We believe it is probable that in the next few years, with the rapid advancements in AI we’re seeing today, compliance officers’ roles will evolve to a more “supervisory” and “final decision” mode, overseeing the work of the AI, and making sure it complies with their processes and policies.

“At Muinmos, in our field of regulatory compliance, we strongly believe in the importance of continuously working at data analysis and cleansing, ensuring that results/ reports given to clients have undergone severe quality testing. Otherwise relying purely on, for example Chat GPT4 (in its current form) in the area of regulatory compliance will open up more compliance risk and regulatory fines.”

Web3 rise

Omnia CEO Cristian Lupascu also remarked on the topic of AI in financial services compliance, “As the amount and complexity of transactional data increases, compliance officers are turning to AI and privacy enhancing technologies to help them identify irregular patterns, detect fraud, and improve customer due diligence with a specific focus on anti-money laundering efforts.”

While AI adoption for compliance has been around for years and continues to grow, Lupascu mentioned that its application in the Web3 space is still new and has ‘enormous potential to facilitate real-time monitoring and analysis of transactions, while also safeguarding data privacy and security.

He concluded, “By automating compliance processes, analyzing blockchain data, and enhancing transparency and accountability, AI in Web3 can be a critical driver in improving compliance. However, the adoption of AI in Web3 is not without challenges, such as the need for regulatory clarity, investment in data quality, and the necessary infrastructure to support AI adoption.”

Copyright © 2023 RegTech Analyst

Enjoyed the story? 

Subscribe to our weekly RegTech newsletter and get the latest industry news & research

Copyright © 2018 RegTech Analyst

Investors

The following investor(s) were tagged in this article.