Artificial Intelligence (AI) in Finance

Table of Contents:

• Application of AI in finance
• Addressing the challenges of AI in finance
• Leveraging modern technologies to make the most of AI adoption
• Use cases for implementing AI for financial institutions

Application of AI in finance

The conversation around Artificial Intelligence (AI), particularly generative AI, continues to accelerate in the ever-evolving financial technology landscape. AI technologies, including machine learning, are used today to address a wide range of different workflows and customer-facing services, from process automation to optimization.

With the advances in AI in finance come risks that financial institutions need to consider.

Addressing the challenges of AI in finance

While the industry has always had to deal with persistent issues like risk management and governance, adopting generative AI and machine learning introduces new challenges that Artificial Intelligence specialists have always dealt with, like inherent biases and ethical concerns. One challenge that stands out for generative AI is hallucination – the generation of content that is not accurate, factual, or reflective of the real world. AI models may produce information that sounds plausible but is entirely fictional.

Generative AI models, especially in natural language processing, might generate text that is coherent and contextually appropriate but lacks factual accuracy. This poses challenges in different domains.

• Misleading financial planning advice: In financial advisory services, hallucinated information may result in misleading advice, leading to unexpected risks or missed opportunities.
• Incorrect risk assessments for lending: Inaccurate risk profiles may lead to poor risk assessments for loan applicants, which can cause financial institutions to approve a loan with a higher risk of default than the firm would normally accept.
• Sensitive information in generated text: When generating text, models may inadvertently include sensitive information from the training data. Adversaries can craft input prompts to coax the model into generating outputs that expose confidential details present in the training corpus.

A strategic and comprehensive approach encompassing various aspects of technology, data, ethics, and organizational readiness is critical to overcoming these challenges when discussing AI in finance.

• Hallucination mitigation: One promising strategy is to use the retrieval augmented generation (RAG) approach to mitigate hallucination in generative AI models – incorporating information retrieval mechanisms to enhance the generation process and ensuring that generated content is grounded in real-world knowledge. MongoDB Atlas Vector Search is an effective mechanism to support RAG, which uses vector search to retrieve relevant documents based on the input query.
• Data quality and availability: Take a step back before adopting AI in finance to ensure the quality, relevance, and accuracy of data being used for AI training and decision-making can be accessed in real time.
• AI education: The key is to invest in training programs to address skill gaps, create a culture of learning and development, and promote awareness about vulnerabilities.
• Develop new governance, frameworks, and controls: Before going live, create safe and secure environments for testing and learning that allow you to fail fast and course-correct.
• Monitoring and continuous improvement: Implement robust monitoring systems to measure and understand financial impacts, scale, and complexity associated with adopting AI in finance.
• Security and privacy: Implement robust cybersecurity measures to safeguard AI models and the data they rely on.

Leveraging modern technologies to make the most of Artificial Intelligence adoption

Financial institutions should not only focus on current product enhancements but also future-proof their capabilities through infrastructure modernization. When adopting advanced technologies like AI and ML, which require data as the foundation, organizations often struggle to integrate these innovations into legacy systems due to their inflexibility and resistance to modification. The following recommendations will help unlock the transformative potential of AI at scale while ensuring privacy and security concerns are addressed:

• Train AI/ML models on the most accurate and up-to-date data addressing the critical need for adaptability and agility. By unifying data from backend systems to customer interactions, financial institutions can surface insights in real time.
• Future-proof with a flexible data schema that accommodates any data structure, format, or source. This flexibility facilitates seamless integration with AI/ML platforms without extensive infrastructure modifications.
• Address security concerns with built-in security controls across all data. Robust security features such as authentication, role-based access controls, and comprehensive data encryption are essential, whether managed in a customer environment or the cloud.
• Launch and scale always-on and secure applications by integrating third-party services with APIs. A flexible data model can handle structured and unstructured data, making it a great fit for orchestrating your open API ecosystem.

Use cases for implementing AI for financial institutions

Emerging use cases for AI in payments

A lack of developer capacity is one of the biggest challenges for banks when delivering payment product innovation. Banks believe the product enhancements they could not deliver in the past two years due to resource constraints would have supported a 5.3% growth in payments revenues. With this in mind and the revolutionary transformation with the integration of AI, it is imperative to consider how to free up developer resources to make the most of these opportunities. Below are some of the opportunities Celent, a leading research and advisory firm for financial institutions, highlights for banks to harness the benefits of AI to improve payment operations.

Relationship management support with chatbots

One key service that relationship managers provide to their private banking customers is aggregating and condensing information. Because banks typically operate on fragmented infrastructure, this can require a lot of detailed knowledge about this infrastructure and how to source information, such as:

• When are the next coupon dates for bonds in the portfolio?
• What has been the cost of transactions for a given portfolio?
• What would be a summary of our latest research?
• How do you generate a summary of a conversation with the client?

Until now, these activities would be highly manual and exploratory. For example, a relationship manager (RM) looking for the next coupon dates would likely have to go into each of the client's individual positions and manually look up the coupon dates. If this is a frequent enough activity, the RM could raise a request for change with the product manager of the portfolio management software to add this as a standardized report. But even if such a standardized report existed, the RM might struggle to find the report quickly. Overall, the process is time-consuming.

Generative Artificial Intelligence systems can facilitate such tasks. Even without specifically trained models, RAG can be used to have the AI generate the correct answers, provide the inquirer with a detailed explanation of how to get to the data, and, in the same cases, directly execute the query against the system and report back the results. The algorithm must have access to not only the primary business data, e.g., the customer portfolio data, but also user manuals and static data. Detailed customer data, in machine-readable format and as text documents, is used to personalize the output for the individual customer.

In an interactive process, the RM can instruct the AI to add more information about specific topics, tweak the text, or make any other necessary changes. Ultimately, the RM will be the quality control for the AI’s output to mitigate hallucinations or information gaps.

As outlined above, not only will the AI need highly heterogeneous data from highly structured portfolio information to text documents and system manuals to provide a flexible natural language interface for the RMs, it will also need timely processing of information about a customer's transactions, positions, and investment objectives. Providing transactional database capabilities and vector search makes it easy to build RAG-based applications using MongoDB’s developer data platform.

Risk management and regulatory compliance

Risk and fraud detection

Banks are tasked not only with safeguarding customer assets but also with fraud detection, verifying customer identities through "know your customer" capabilities (KYC), supporting sanctions regimes, and preventing various illegal activities through anti-money laundering (AML). The challenge is magnified by the sheer volume and complexity of regulations, making integrating new rules into bank infrastructure costly, time-consuming, and often inadequate. AI offers a transformative approach to fraud detection and risk management by automating the interpretation of regulations, supporting data cleansing, and enhancing the efficacy of surveillance systems. Unlike static, rules-based frameworks that may miss or misidentify fraud due to narrow scope or limited data, AI can adaptively learn and analyze vast datasets to identify suspicious activities more accurately. Machine learning, in particular, has shown promise in trade surveillance, offering a more dynamic and comprehensive approach to fraud detection.

Regulatory compliance and code change assistance

Traditionally, adapting to new regulations has required the manual translation of legal text into code, provisioning of data, and thorough quality control – a costly and time-consuming process, often leading to incomplete or insufficient compliance. Artificial Intelligence has the potential to revolutionize compliance by automating the translation of regulatory texts into actionable data requirements and validating compliance through intelligent analysis. This approach is challenging, as AI-based systems may produce non-deterministic outcomes and unexpected errors. However, the ability to rapidly adapt to new regulations and provide detailed records of compliance processes can significantly enhance regulatory adherence.

Financial document search and summarization

Financial institutions handle a broad spectrum of documents critical to their operations. For example, retail banks focus on contracts, policies, credit memos, underwriting documents, and regulatory filings, which are pivotal for daily banking services. On the other hand, capital market firms delve into company filings, transcripts, reports, and intricate data sets to grasp global market dynamics and risk assessments.

These documents often arrive in unstructured formats, presenting challenges in efficiently locating and synthesizing the necessary information. Both retail banks and capital market firms allocate considerable time to searching for and condensing information from documents internally, resulting in reduced direct engagement with their clients.

Generative AI can streamline finding and integrating information from documents by using natural language processing (NLP) and machine learning to understand and summarize content. This reduces the need for manual searches, allowing staff to access relevant information more quickly.

ESG analysis

The profound impact of environmental, social, and governance (ESG) is evident, driven by regulatory changes, especially in Europe, compelling the integration of ESG into investment and lending decisions. Regulations such as the EU Sustainable Finance Disclosure Regulation (SFDR) and the EU Taxonomy Regulation are examples of such directives that require financial institutions to consider environmental sustainability in their operations and investment products. The regulatory and commercial requirements in turn, drive banks also to improve their green lending practices. This is a strategic shift for the finance industry, attracting clients, managing risks, and creating long-term value.

However, the finance industry faces many challenges when improving its ESG analysis, including defining and aligning standards and processing and managing the flood of rapidly changing and varied data to be included for ESG analysis purposes.

AI can help to address these key challenges in not only an automatic but also adaptive manner via techniques like machine learning. Financial institutions and ESG solution providers have already leveraged AI to extract insights from corporate reports, social media, and environmental data, improving the accuracy and depth of ESG analysis. As the market demands a more sustainable and equitable society, predictive AI combined with generative AI can also help to reduce bias in lending to create fairer and more inclusive financing while improving predictive powers. The power of AI can help facilitate the development of sophisticated sustainability models and strategies, marking a leap forward in integrating ESG into broader financial and corporate practices.

MongoDB's dynamic architecture revolutionizes ESG data management, handling semi-structured and unstructured data. Its flexible schema allows the adaptation of data models as ESG strategies evolve. Advanced text search capabilities efficiently analyze vast semi-structured data for informed ESG reporting. Support for vector search enriches ESG analysis with multimedia content insights.

Below is a diagram of an enterprise ESG solution architecture with the boxes labeled with the green leaf where MongoDB can be deployed to support the ESG data analytics-related services.

Transforming credit scoring with AI for the financial services industry

The convergence of alternative data, Artificial Intelligence, and generative AI is reshaping the foundations of credit scoring, marking a pivotal moment in the financial industry. The challenges of traditional models are being overcome by adopting alternative credit scoring methods, offering a more inclusive and nuanced assessment. Generative AI, while introducing the potential challenge of hallucination, represents the forefront of innovation, not only revolutionizing technological capabilities but fundamentally redefining how credit is evaluated, fostering a new era of financial inclusivity, efficiency, and fairness.

The use of Artificial Intelligence, in particular Generative Artificial Intelligence, as an alternative method to credit scoring has emerged as a transformative force to address the challenges of traditional credit scoring methods for several reasons:

• Alternative data analysis: Unlike traditional models that rely on predefined rules and historical credit data, AI models can process a myriad of information, including alternative data such as utility payments and rental history, to create a more comprehensive assessment of an individual's creditworthiness, ensuring that a broader range of financial behaviors is considered.
• AI offers unparalleled adaptability: As economic conditions change and consumer behaviors evolve, AI-powered models can quickly adjust and learn from new data. This continuous learning ensures that credit scoring remains relevant and effective in ever-changing financial landscapes.
• Fraud detection: AI algorithms can detect fraudulent behavior by identifying anomalies and suspicious patterns in credit applications and transaction data.
• Predictive analysis: AI algorithms, particularly ML techniques, can be used to build predictive models that identify patterns and correlations in historical credit data, forecasting the greater likelihood of loan defaults.
• Behavioral analysis: AI algorithms can analyze behavioral data sets to understand financial habits and risk propensity. By monitoring real-time financial behavior, AI models can provide dynamic credit scores that reflect current risk profiles.

By harnessing the power of Artificial Intelligence, lenders can make more informed lending decisions, expand access to credit, and better serve consumers (especially those with limited credit history). However, to mitigate potential biases and ensure consumer trust, it's crucial to ensure transparency, fairness, and regulatory compliance when deploying Artificial Intelligence in credit scoring.

Other notable use cases

• Risk modeling: AI can create synthetic scenarios and data that can be used to stress test financial systems and models.
• Personalized wealth management: The integration of AI empowers institutions to offer personalized advice and solutions. By analyzing vast data sets, AI can help provide valuable insights for making informed decisions and optimizing investment portfolios. Wealth managers can customize investment strategies to individual preferences, risk tolerance, and financial goals.
• Algorithmic trading: AI algorithms can analyze public market data and execute trades at speed, optimizing trading strategies.
• Generating financial reports: AI can analyze financial data including transactions, invoices, and account statements, to automate reports being generated. By using AI and ML techniques, relevant information can be extracted where required.

These examples highlight several avenues for integrating AI within financial institutions. Embracing AI in financial applications promises enhanced risk management, operational efficiency, and superior customer experiences. Therefore, it is essential for financial institutions to grasp the profound technological implications, scale, and intricacies associated with AI, particularly in crafting a generative AI strategy. Adopting a strategic and holistic approach that addresses technological, data, ethical, and organizational dimensions is imperative for navigating this transformative landscape effectively.