ATOM Consulting Series 1

Opportunities & Implications

In this series, we delve into the transformative landscape of financial services, exploring the opportunities and challenges presented by ongoing digitalization.


We examine key trends, including Technology (AI, Web3.0), Regulation and Risk management, through the lens of strategy, compliance, and organisational culture. 


By interviewing industry experts and clients, we provide valuable insights into this evolving landscape and help businesses navigate the complexities of the digital age.

ATOM Consulting Series 2

Productivity in Banking and Wealth Management

A closer look at how AI is driving productivity and the solutions Institutions are deploying.


In this second of the series, we take a deeper dive into how AI is boosting productivity and reshaping operations in banking and wealth management.  Looking at the benefits Institutions are deriving and the types of AI being deployed. 

ATOM Consulting Series 3    Updated 10/10/24

AI Regulation: A Global Perspective

We delve into the varying regulatory approach to AI Deployment


In this third of the series, we examine the key regulatory trends worldwide as jurisdiction seek to ensure these powerful technologies are used responsibly and ethically.

The Future of Finance

AI Transforming the Future of Finance

Artificial Intelligence (AI) is rapidly reshaping the financial services industry, driving productivity gains and enhancing customer experiences across banking and wealth management. From automating routine tasks to delivering personalised financial advice, the applications of AI are far-reaching and gaining significant traction. Since we first took a look at the key trends for AI for financial services the market has developed significantly.  We are revisiting this boardroom topic and looking at what businesses should be prioritising as they prepare for the future.


AI Takes Center Stage in Financial Services

Leading banks and wealth management firms are increasingly turning to AI to streamline operations and stay competitive. For example, JPMorgan Chase has deployed AI-powered chatbots to handle routine customer inquiries, freeing up human agents to focus on more complex matters. Similarly, UBS has integrated robo-advisory capabilities powered by AI to provide personalised investment recommendations to its clients.

Beyond customer-facing applications, AI is also proving invaluable in back-office functions. Santander Bank has implemented AI-driven document processing to automate data extraction from financial statements, dramatically reducing manual effort. And in the realm of risk management, HSBC utilises AI-based fraud detection systems to identify suspicious transactions in real-time.

Regulators Grapple with AI Governance

As AI proliferates in finance, regulators around the world are working to establish guidelines and frameworks to ensure its responsible use. In the United States, the Securities and Exchange Commission (SEC) has emphasised the importance of transparency and oversight when it comes to AI-driven financial advice. "Firms must be able to explain how their AI systems work and how they're making decisions," said SEC Commissioner Hester Peirce.


Across the Atlantic, the European Union has taken a more proactive approach, with the proposed AI Act seeking to introduce comprehensive regulations for AI applications, including those in the financial sector. "We want to make sure that AI is developed and used in a way that respects our values and fundamental rights," explained Margrethe Vestager, Executive Vice-President of the European Commission.

In the United Kingdom, the Financial Conduct Authority (FCA) has acknowledged the transformative potential of AI while also highlighting the need for robust governance. "AI has the power to revolutionise financial services, but only if it is implemented responsibly and with appropriate safeguards," said Nikhil Rathi, Chief Executive of the FCA.

The Future of Finance is Intelligent

As the adoption of AI in banking and wealth management continues to accelerate, it is clear that the future of finance will be profoundly shaped by this transformative technology. By embracing AI's capabilities while navigating the evolving regulatory landscape, financial institutions can unlock new levels of efficiency, innovation, and customer-centricity.

In this series, we delve deeper into the specific ways AI is revolutionising the financial services industry, the challenges this is creating, and we explore case studies and insights from industry experts. What does this mean for Financial Services participants, how quickly do you need to respond and the strategic prioritisation steps to optimise investment. 

Productivity in Banking and Wealth Management

In our previous article, we introduced the transformative potential of Artificial Intelligence (AI) in the financial services sector. In this paper we take a deeper dive  into how AI is boosting productivity and reshaping operations in banking and wealth management. Looking at the benefits Institutions are deriving and the types of AI being deployed.


Customer Service Automation: 24/7 Support at Your Fingertips


The days of waiting on hold for a human representative are becoming a thing of the past. AI-powered chatbots and virtual assistants are revolutionising customer service in the financial industry.


Bank of America's virtual assistant, Erica, has become a cornerstone of their customer service strategy. Launched in 2018, Erica has handled over 1 billion client interactions, providing instant responses to account queries, transaction details, and even offering proactive financial advice. Erica utilises Natural Language Processing (NLP) and Machine Learning (ML) algorithms, likely based on a customised Language Model. This sophisticated AI has achieved a 77% first-contact resolution rate and a 90% containment rate, effectively reducing call volume to human agents by 30%.


Similarly, Wells Fargo's AI-powered chatbot can understand natural language queries, allowing customers to check balances, transfer funds, and get answers to frequently asked questions without human intervention. The chatbot employs NLP and predictive analytics, combining rule-based systems with ML models. This implementation has led to a 25% reduction in customer service costs and slashed average response times from 51 seconds to just 5 seconds, all while boosting customer satisfaction scores by 15%.


Personalised Financial Advice: AI-Driven Investment Recommendations


Robo-advisors and AI-driven investment platforms are democratising access to personalised financial advice. These systems analyse vast amounts of data to provide tailored investment recommendations based on an individual's financial goals, risk tolerance, and market conditions.


Betterment, a pioneer in the robo-advisory space, uses AI algorithms to create and manage diversified portfolios for its clients. The platform automatically rebalances portfolios and implements tax-loss harvesting strategies, tasks that would be time-consuming for human advisors. Betterment employs ML algorithms for portfolio optimisation and Reinforcement Learning for continuous improvement of investment strategies. This AI-driven approach has led to impressive results, with 88% of Betterment portfolios outperforming their benchmarks over a 4-year period, and their automated tax-loss harvesting adding an estimated 0.77% to annual investment performance.


Wealth management giant BlackRock has integrated AI into its Aladdin platform, which now offers personalised portfolio construction and risk management tools to financial advisors. Aladdin utilises a combination of predictive analytics, ML, and recently, aspects of Generative AI for scenario generation. This AI-augmented approach allows advisors to provide more sophisticated, data-driven advice to their clients, resulting in a 23% improvement in portfolio manager productivity and 25% more accurate risk forecasts compared to traditional models.


Risk Assessment and Fraud Detection: AI as the Financial Watchdog


AI's ability to process vast amounts of data in real-time makes it an invaluable tool for risk assessment and fraud detection in the financial sector.


In credit scoring, AI models are being used to assess creditworthiness more accurately than traditional methods. For instance, Lenddo, a Singapore-based company, uses AI to analyse non-traditional data points like social media activity and smartphone usage patterns to determine credit scores for individuals with limited credit history. Their ML algorithms, which include ensemble methods like Random Forests and Gradient Boosting, along with NLP for text analysis, have increased approval rates for thin-file customers by 50% while reducing default rates by 12% compared to traditional credit scoring methods.


For fraud detection, HSBC has implemented an AI system that has helped reduce fraudulent transactions by 50%. The system analyses customer behavior patterns and flags unusual activities in real-time, allowing for immediate intervention. It uses a combination of Supervised and Unsupervised ML algorithms, including Anomaly Detection models and Deep Learning networks, processing over 1 million transactions per second in real-time.


Mastercard's Decision Intelligence uses AI to analyse various data points during transactions, significantly reducing false declines while maintaining high fraud detection rates. The system employs a sophisticated ML pipeline, including real-time feature engineering and an ensemble of predictive models. This has resulted in a 40% reduction in false declines and approved an additional $2 billion in transactions annually, all while improving fraud detection accuracy by 50%.


Process Automation: Streamlining Back-Office Operations


AI is not just transforming customer-facing operations; it's also revolutionising back-office processes in financial institutions.


JPMorgan Chase has implemented a system called COiN (Contract Intelligence) that uses AI to review commercial loan agreements. This system can accomplish in seconds what previously took lawyers and loan officers 360,000 hours annually. COiN utilises NLP and ML for document understanding, likely based on a customised Large Language Model (LLM). It achieves 99% accuracy in extracting 150 attributes from complex legal documents, reducing processing time for certain documents from 30 minutes to mere seconds.


In document processing, AI-powered Optical Character Recognition (OCR) technology is being used to automate data extraction from various financial documents. For example, ING Bank has deployed AI to process trade finance documents, reducing processing time from days to hours. Their system combines OCR with NLP and ML models for document classification and information extraction, resulting in an 80% reduction in manual data entry errors and enabling a 40% increase in trade finance transaction volume without additional staff.


Robotic Process Automation (RPA) combined with AI is also being used to automate repetitive tasks. UBS has implemented RPA for various back-office processes, resulting in significant time savings and reduced error rates. By integrating traditional RPA with ML models for decision-making and adaptability, UBS has achieved an 85% reduction in processing time for certain tasks, 30% cost savings in back-office operations, and a 90% reduction in data entry errors.


The AI-Powered Future of Finance


As we've seen, AI is driving productivity gains across various aspects of banking and wealth management. From providing 24/7 customer support to offering personalised financial advice, from enhancing risk management to streamlining back-office operations, AI is reshaping the financial services landscape.


Here's a summary of the case studies covered:


Company
AI Application
Type of AI
Performance Benefits
Bank of America (Erica)
Customer Service
NLP, ML, Custom LM
    77% first-contact resolution,  30% reduction in call volume
Wells Fargo
Customer Service
NLP, Predictive Analytics
    25% cost reduction, 90% faster response time
Betterment
Robo-Advisory
ML, Reinforcement Learning
    0.77% added annual performance, 88% portfolios outperforming benchmarks
BlackRock (Aladdin)
Investment Management 
Predictive Analytics, ML, Generative AI 
    23% improved productivity 25% more accurate risk forecasts
Lenddo
Credit Scoring
ML (Random Forests, Gradient Boosting), NLP
    50% increase in thin-file approvals 12% reduction in defaults
HSBC
Fraud Detection
Supervised & Unsupervised ML, Deep Learning 
    50% reduction in fraudulent transactions 
Mastercard
Transaction Processing
ML Pipeline, Ensemble Models
    40% reduction in false declines, $2B additional approved transactions
JPMorgan Chase (COiN)
Document Review 
NLP, ML, Custom LLM
    360,000 hours saved annually 99% accuracy in data extraction
ING Bank
Document Processing
OCR, NLP, ML 

    Processing time reduced from days to hours 80% reduction in errors

UBS
Back-Office Automation
RPA with ML
    85% reduction in processing time 30% cost saving


As AI continues to evolve, we can expect even more sophisticated applications in the financial sector. The type of AI deployed and the extent it is integrated and customised will be use case dependant. However, with these advancements come challenges as customer data becomes a more valuable commodity in the development of models, privacy and ownership policies will need to evolve and adapt. In our next article, we'll explore how regulators are addressing the governance of AI in the financial industry, ensuring that these powerful tools are used responsibly and ethically.

AI Regulation: A Global Perspective

As Artificial Intelligence (AI) and Generative AI (GenAI) continue to transform the financial services industry, regulators worldwide are grappling with how to ensure these powerful technologies are used responsibly and ethically. Key global trends include a focus on explainability and transparency of AI models, emphasis on fair lending and avoiding bias, push for robust governance and risk management frameworks, and a growing concern for data protection and privacy. However, the approach to these issues varies significantly across jurisdictions, with some favouring prescriptive rules and others opting for principles-based guidance.

Regulatory Landscape


United States: Balancing Innovation with Consumer Protection

In the U.S., regulators are taking a cautious but proactive approach, focusing on explainability, fairness, and robust governance. They are striving to balance the promotion of innovation with the need for consumer protection and financial stability.

    The Federal Reserve, FDIC, and OCC jointly released guidance on model risk management for AI/ML systems in financial institutions (2021). This guidance emphasises the need for robust validation and monitoring of AI models, particularly those used in credit decisioning. 

    In June 2023 FIDC and OCC produced further guidance on Banking’s use of data and external providers (AI Models).

    The SEC proposed new rules requiring disclosure of AI use in investment strategies (2022). These rules aim to increase transparency for investors and ensure that AI-driven strategies are subject to appropriate oversight.  The SEC continues to provide guidance and in May 2024 finalised amendments to Regulation S-P governing privacy as part of the Fair and Accurate Credit Transactions Act of 2003.

    The CFPB issued a circular on adverse action notices when using complex AI/ML models in credit decisions (2022). This guidance underscores the importance of providing clear explanations to consumers, even when using complex AI models.


While there appears to be no coordinated regulatory approach the Whitehouse and Senate are taking a proactive approach to shape the direction of US regulation through:

The Biden Executive Order 14110 (October 2023)

  • This focuses on: Safety, security, innovation, worker support, AI bias and civil rights, consumer protection, privacy, federal AI usage, and international leadership, aiming to promote US leadership in the AI space.
  • The Executive order covers 50 federal entities promoting them to take more than 100 actions to implement the guidance.
  • Covers enforcement of existing regulations, and development of guidance around AI.


Roadmap for Artificial Intelligence Policy (May 2024)

A bipartisan initiative that through policy guidance aims to: support U.S. AI innovation, AI workforce impacts, identify high-impact AI uses, protect elections, privacy, transparency, intellectual property, and identify AI risks.


For Financial services this means:

  • Development of a comprehensive federal data privacy framework for AI across sectors.
  • Provisions for data minimisation, security, consumer rights, consent, disclosure, and data brokers.
  • Accurate and representative data in AI models for financial service providers.
  • Regulatory gap analysis in the financial sector.


Overall, both the Executive Order and the Roadmap aim to promote US involvement in AI and to establish frameworks for safe, secure, and trustworthy AI development.


Europe: Comprehensive and Risk-Based Regulation

The EU has taken a comprehensive, risk-based approach to AI regulation, with a strong emphasis on data protection, privacy, and human oversight. This approach is exemplified by the AI Act which came into force in August 2024, and sets a new global standard for AI regulation.

The AI Act classifies AI systems in financial services as "high-risk," requiring stringent oversight and by 2025 will prohibit the use of certain Ai systems. This analysis and classification means that AI systems used in credit scoring, insurance pricing, and other critical financial functions will be subject to rigorous requirements for accuracy, robustness, and transparency.


The European Insurance and Occupational Pensions Authority (EIOPA) released guidelines on the use of AI in insurance (2023). These guidelines address the specific challenges of using AI in insurance, including fairness in pricing and claims handling.


The European Banking Authority (EBA) published guidelines on the use of ML for internal ratings-based models (2022). These guidelines provide a framework for banks to incorporate ML into their credit risk models while maintaining compliance with existing capital requirements. 



                            Asia: Fostering Innovation with Responsible Governance


                            Asian countries have adopted varied approaches to AI regulation in finance, with a general trend towards fostering innovation while ensuring responsible governance. There's a strong focus on customer protection and cross-border collaboration.

                            Singapore's Monetary Authority (MAS) issued guidelines on the use of AI and data analytics in financial services (2022). These principles-based guidelines emphasise fairness, ethics, accountability, and transparency (FEAT) in the use of AI.


                            Hong Kong's Monetary Authority (HKMA) introduced a bank AI supervision framework (2023). This framework adopts a risk-based approach, focusing on consumer protection and the responsible development of AI in banking.


                                Japan's Financial Services Agency (FSA) published principles for AI governance in the financial sector (2022). These principles emphasise the importance of human-centric AI and the need for ongoing monitoring and adjustment of AI systems. 


                                  China with its Government led initiative has adopted a broad set of regulations and guidance around AI.  Including: Interim Administrative Measures for Generative Artificial Intelligence Services: Comprehensive regulation for generative AI, covering areas like algorithm transparency, data security, and content control. It is also promoting the development of standards through technical guidelines for AI development and application, such as the "General Requirements for Information Security Technology—Security Guidelines for Generative AI.” 2023


                                  Australia: Principles-Based and Consumer-Focused

                                  Australia has taken a principles-based approach to AI regulation in finance, with a strong focus on accountability, governance, and consumer protection.  The direction is coming from the government more so than regulators with the 2023/24 budget including funding to support AI Adoption.  Key guidance comes from the 2023 reports: Safe and Responsible AI in Australia  and Rapid Response Information Report: Generative AI.  Financial Services specific regulatory is found in the earlier releases from ASIC and APRA:  

                                    The Australian Securities and Investments Commission (ASIC) released guidance on the use of AI in financial services and credit activities (2022). This guidance emphasises the importance of accountability and transparency in AI-driven decision-making.

                                    The Reserve Bank of Australia (RBA) and Australian Prudential Regulation Authority (APRA) jointly issued a paper on ML use in financial services (2023). This paper highlights the potential benefits and risks of ML in finance and outlines expectations for its responsible use.


                                  Case Studies


                                  We examine some notable use cases of AI in financial services:

                                  Risk Assessment and Compliance Testing: JPMorgan Chase's COIN

                                  JPMorgan Chase's Contract Intelligence (COIN) system demonstrates the power of AI in streamlining legal and compliance processes. COIN uses Natural Language Processing (NLP) and Machine Learning (ML) to analyse legal documents and extract relevant information.

                                  The system can review commercial loan agreements in seconds, a task that previously took lawyers and loan officers 360,000 hours annually. COIN achieves 99% accuracy in extracting 150 attributes from complex legal documents, significantly reducing the risk of human error and freeing up skilled professionals for higher-value tasks.

                                  COIN's success has led to its expansion across various departments at JPMorgan Chase, including handling routine IT requests and analysing corporate communications for compliance issues. This case illustrates how AI can dramatically improve efficiency and accuracy in risk assessment and compliance testing.


                                  Credit Assessment: Upstart's AI-Powered Lending Platform

                                  Upstart has revolutionised personal lending with its AI-powered platform that uses alternative data and machine learning to assess creditworthiness. Unlike traditional credit scoring models that rely heavily on FICO scores, Upstart's model considers over 1,000 variables, including education, job history, and even how an applicant interacts with the loan application.

                                  The results have been impressive: 75% of loans through Upstart's platform are fully automated, requiring no human intervention. More importantly, the platform has achieved 53% lower loss rates compared to traditional models while approving more applicants. This has made credit more accessible to underserved populations, including younger applicants and those with limited credit histories.

                                  Upstart's success demonstrates the potential of AI to make credit assessment more inclusive and accurate, aligning with regulatory goals of fair lending and financial inclusion.


                                  Fraud Detection: Danske Bank's AI System

                                  Danske Bank, Denmark's largest bank, implemented an AI-based fraud detection system to combat the rising tide of financial fraud. The system uses machine learning algorithms in conjunction with rule-based systems to analyse transactions in real-time.

                                  The AI model considers over 200 factors for each transaction, learning and adapting to new fraud patterns over time. This dynamic approach has led to a 60% reduction in false positives and a 50% increase in the fraud detection rate. For customers, this means fewer legitimate transactions are flagged as suspicious, leading to a smoother banking experience.

                                  Danske Bank's case highlights how AI can enhance fraud detection capabilities while improving customer experience, addressing the regulatory concerns of both financial security and consumer protection.


                                  Onboarding, KYC, and AML: HSBC's AI-Powered KYC Solution

                                  HSBC has deployed an AI-powered Know Your Customer (KYC) solution to streamline customer onboarding and ongoing due diligence processes. The system uses a combination of Natural Language Processing, Machine Learning, and Optical Character Recognition (OCR) to automate the collection and analysis of customer information.


                                  This AI solution has reduced customer onboarding time by 80% and achieved 70% cost savings in KYC processes. It can quickly scan and analyse vast amounts of data from various sources, including news articles and social media, to identify potential risks associated with customers.


                                  HSBC's KYC solution demonstrates how AI can enhance regulatory compliance while improving operational efficiency and customer experience. It addresses the regulatory requirements for robust AML procedures while also aligning with the trend towards faster, more digital banking services.


                                  Regulatory Challenges and Future Outlook

                                  As AI continues to evolve, regulators face several challenges:

                                  • Keeping pace with innovation: The rapid development of AI, especially GenAI, makes it difficult for regulations to stay current.
                                  • Balancing innovation and risk: Regulators must foster innovation while ensuring financial stability and consumer protection.
                                  • Cross-border harmonisation: With financial services increasingly global, there's a need for international coordination on AI regulation.
                                  • Explainability vs. performance: Some high-performing AI models (e.g., deep learning) can be difficult to explain, creating tension with transparency requirements.
                                  • Data privacy and security: As AI systems rely on vast amounts of data, ensuring privacy and security becomes increasingly complex.


                                  Looking ahead, we can expect:

                                        • More specific guidance on the use of GenAI in financial services
                                        • Increased focus on AI auditing and algorithmic impact assessments
                                        • Development of AI-specific risk management frameworks
                                        • Greater emphasis on AI ethics and responsible AI practices
                                        • Potential creation of regulatory sandboxes for testing AI innovations


                                  As the technology and regulatory landscape continues to evolve, financial institutions must stay informed and proactive in their approach to AI governance. An effective Digital Strategy is not be about innovation alone,  by incorporating regulatory compliance into the future operating models and  embracing responsible AI practices, the financial services industry can leverage the power of AI and Digital Infrastructure to create new highly efficient business models, that are compliant by design.