Banking and financial services are among the most heavily regulated industries, by a plethora of local and global regulatory bodies across jurisdictions. Yet financial crimes across the globe have steadily risen in the last two decades – not just in volumes, but also in terms of complexity and sophistication!
As the cost of compliance continues to remain very high – estimated to be over $200bn annually for KYC-AML alone, reducing compliance cost while improving operational efficiency in this function at banks and financial institutions (FIs) are among the top priorities of CXOs across the globe!
In their pursuit of enhanced compliance, FIs have started looking at data driven frameworks, requiring massive volumes of data processing for risk based decisioning. As artificial intelligence (AI) powered advanced technology solutions started emerging with unmatched complex data analysis and pattern detection capabilities, FIs began assessing the value that these could bring to Regulatory compliance.
In this paper, we explore how AI is revolutionizing banking compliance, and the future roadmap of AI powered digitization in this space.
Banking Compliance – The Dilemma of Balancing AI and Regulatory Expectations
For the financial sector, Know Your Customer (KYC) and Anti-Money Laundering (AML) compliance form the foundational pillars to provide any financial service. Prevention and detection of fraud and other financial crimes are also key functions of FIs. Over the years, Regulations around these functions have grown more stringent and complex, while also expanding in breadth and depth of coverage.
For FIs, manual processes have given way to technology based compliance, which are primarily rule-driven. These systems have clear audit trails, and the rules are explainable to Regulators. However, the growing scale of financial crimes are testimony to the fact that these systems are not effective or efficient enough in combating today’s sophisticated criminal mechanisms!
AI has been rapidly changing everything around us, in almost every aspect of our lives. From online shopping to entertainment, ride hailing apps to food orders, virtual assistants to driverless cars, and so much more – all these are being powered by AI! FIs however have been cautious about AI adoption particularly in the compliance space.
The reason largely has been due to uncertainties around Regulators’ acceptance of machine based decisioning, explainability of self-learning algorithms, fairness of machine learning logic in identifying suspicious activities and some others.
AI Adoption in Banking Compliance – The Journey So Far
In the world of banking compliance, AI based decisioning was viewed with scepticism as most machine learning models were black box in nature. This made it difficult to explain the basis of decisions taken by the algorithms while detecting fraudulent or suspicious customer behaviour, and FIs were not confident of evidencing the same to Regulators. Consequently, AI adoption by FIs in the compliance area remained low and patchy.
Regulators across the world have more recently started acknowledging the virtues of AI, specially its power to process colossal amounts of data – structured and unstructured – with high speed and accuracy, to generate meaningful insights.
From 2018 onwards, leading Regulators like FinCEN (USA), European Commission (EU), FCA (UK), MAS (Singapore), AUSTRAC (Australia) and others have been supporting and encouraging FIs within their jurisdiction to leverage AI to improve efficiency and effectiveness of their financial crime compliance programs.
Support from Regulators also came in the form of digital sandboxes, where FIs and FinTech players can experiment with AI and advanced technology based innovations for KYC-AML and other compliance functions. This has accelerated development of niche AI powered fincrime compliance solutions, which witnessed a spurt during COVID-19 when all financial services moved online.
Fincrime compliance has traditionally been a manual intensive and human judgement driven function. So the debate around man versus machine in this critical area continues, as many argue that AI can be ‘augmented intelligence’ for humans to execute this function more effectively, efficiently.
This rationale got a shot in the arm when several fintechs with fully AI based automated KYC-AML processes became victims of very high rate of frauds (around 75%) while onboarding new customers and disbursing Govt relief packages during COVID-19.
FATE of AI – Bringing Trust to Digital Compliance Solutions
The adoption and success of any AI solution in the world of banking compliance, irrespective of how good its output is, rests to a large extent on its FATE – fairness, accountability, transparency and explainability. Fairness is to ensure that the AI models make unbiased decisions or predictions, i.e. no biases around gender, ethnicity, race, religion, health, economic background among others.
Biases are likely to creep in AI algorithms through training data which include decisions made by humans during manual compliance processes. A fair solution should be able to identify biases in data and models and remediate them accordingly.
Accountability, transparency and explainability are must-haves in any AI based compliance solution to understand the outcome – decision or prediction – making it auditable for the FIs as well as Regulators.
The FI must be able to demonstrate the outcomes, e.g. why and how a certain activity was considered fraudulent or suspected of a financial crime or otherwise. This is to ensure that FIs’ compliance policies and Regulatory mandates are being adhered to during decision making, moving away from a black box approach, thus enabling auditability.
In April 2021, European Commission, EU’s Regulatory authority proposed new rules on trustworthy AI, setting in motion the world’s first legal framework on AI. Following a risk based approach, AI solutions will be classified into minimal, limited, high or unacceptable risk through scoring on several parameters including trust, security, inclusivity and transparency among others.
Such Regulatory guidance is bound to boost the use of AI in compliance, the minimal and limited risk ones with the highest confidence!
Trustworthy AI for Banking Compliance Digitization – The Road Ahead
AI can be leveraged in several critical banking compliance functions e.g. KYC during new customer onboarding; customer lifecycle management and ongoing due diligence; risk scoring; preventive and detective controls around money laundering, fraud, Sanctions, bribery & corruption etc.; suspicious network detection; alert investigation; reporting and so much more!
As Regulatory acceptance of AI in compliance has been growing, with options of co-development and certification of AI based solutions by Regulators, this initiative will certainly witness a spurt in adoption by FIs.
We may very soon see AI driven platforms for end-to-end banking compliance functions, as machine learning and analytics work with RPA and chatbots, alongside tools like OCR (optical character recognition) and OCR confidence reporting engines to complete the automation landscape.
But a path-breaking application of AI would be to build a solution that can pre-empt what financial criminals might do to circumvent the prevention and detection mechanisms of FIs to commit financial crimes and create algorithms accordingly for blocking such crimes.
Can we expect AI to create its own threat intelligence list and generate alerts when transactions involving such individuals/entities are initiated? Can NLP be built to interpret and analyse new regulations, and recommend changes to be implemented in the bank’s systems and processes as a result? This vision may become a reality soon, as AI’s journey in banking compliance has just begun….
