What is AI in banking?

To stay ahead of technology trends, increase their competitive advantage, and provide valuable services and better customer experiences, financial services firms like banks have embraced digital transformation initiatives.

The advent of AI technologies has made digital transformation even more important, as it has the potential to remake the industry and determine which companies thrive.

Historically, incumbent financial service providers have struggled with innovation. A McKinsey study¹ found that large banks were 40% less productive than digital natives. Many emerging banking startups are pioneering artificial intelligence use cases, making it even more important that traditional banks catch up and innovate themselves.

Investment banking firms have long used natural language processing (NLP) to parse the vast amounts of data they have internally or that they pull from third-party sources. They use NLP to examine data sets to make more informed decisions around key investments and wealth management.

The banking sector, specifically, is absorbing the desired benefits of AI technologies. Customers want digital banking experiences: apps where they can learn more information about provided services, interact with people or virtual assistants, and better manage their finances. Companies need to improve the user experience to keep those customers happy. Adopting and deploying AI solutions is one way to accomplish that.

While AI is powerful on its own, combining it with automation unlocks even more potential. AI-powered automation takes the intelligence of AI with the repeatability of automation. For example, AI can enhance robotic process automation (RPA) to better parse data analytics and take actions based on what the AI decides is best. One example is banks that use RPA to validate customer data needed to meet know your customer (KYC), anti-money laundering (AML) and customer due diligence (CDD) restrictions.

Financial services organizations are embracing artificial intelligence (AI) for various reasons, such as risk management, customer experience and forecasting market trends.

AI helps customers enhance their decision-making about financial matters. They are more likely to stay with banks that use cutting-edge AI technology to help them better manage their money.

But given extensive industry regulations, banks and other financial services organizations need a comprehensive strategy for approaching AI. Using AI requires a thoughtful framework to mitigate risk and exposure.

The IBM Institute for Business Value published a guide for banks looking to embed AI tools and practices into their operations in its 2024 Global Outlook for Banking and Financial Markets report. Some of the key actions are:

• Define the AI governance and risk profile of the bank: Every bank is different, and the leaders of each bank must make its own decisions about AI risk and deployment. Banks should embrace AI with the knowledge that it requires them to counteract any potential risks with strong security measures.
• Prioritize use cases: AI deployments must be tied to specific business use cases that drive measurable impact and align to organizational goals. Examples of specific use cases are customer-facing chatbots, personalized investment strategies, fraud prevention and creditworthiness scoring.
• Choose a trusted AI platform: Most enterprise AI approaches require the application of multiple AI models to ensure that an organization has everything it needs to succeed. Therefore, banks need to choose whether to use open-source models, models built in-house, or both.
• Embrace a hybrid cloud architecture: AI requires banks to address any existing technology inefficiencies they might have and prioritize application resource management. By using a hybrid cloud architecture, banks can switch between public clouds and private clouds to promote resilience and responsiveness for real-time digital banking.
• Learn from initial deployments: Banks concerned about risks should implement small-scale tests and use cases to assess the impacts before scaling and deploying new implementations. Early lessons are valuable because they help banks better understand what other infrastructure they need to deploy and where they need to make adjustments.
• Create an “AI factory”: Once an organization has established a workable strategy for building or adopting AI for specific use cases, it should build an apparatus that adds AI to its operations and makes it central to all development and business methods.

There are several key benefits for banks that embrace and deploy AI.

• Enhanced cybersecurity and fraud detection: Cyberattackers increasingly use AI to create more sophisticated ways to defraud financial institutions. They can use AI-created audio² to imitate customers, confusing customer service agents. They can use AI to make phishing emails look increasingly legitimate. As a result, those financial institutions need to use AI algorithms to protect their employees from cybersecurity threats in real-time, while creating tools to help customers avoid the same tricks. Financial institutions and governmental agencies can also use AI systems to thwart other financial crimes like money laundering or impersonation.
• Enhanced APIs: Banking operations increasingly depend on the use of application programming interfaces (APIs) to enable customers to track their money on various applications. For example, banks must give API permission to third-party budgeting apps so customers can monitor multiple bank accounts. AI enhances API usage by enabling more security measures and automating repetitive tasks, making them more powerful.
  
• Embeddable banking: This is the introduction of banking into nontraditional experiences, such as when Starbucks started its own payments app³. Embeddable banking is expected to grow as a service, especially as AI helps retailers and other companies collect and analyze data about potential market opportunities, predict creditworthiness, and better personalize services to customers.
• More intelligent customer tools: The rise of generative AI powered by deep learning means that the investment and banking industries can deploy more sophisticated tools to streamline customer service. AI-powered chatbots and virtual assistants can enhance customer support, helping customers solve small problems on their own. AI can also power budgeting apps that help customers better manage their finances and save more money.
• New markets and opportunities: They also use AI for predictive analytics to have better insights into their customers. AI-driven predictive analytics can identify new areas of growth for their business and their customers and can better estimate which customers are a churn risk. For example, banks can analyze their customers’ habits, such as how often they log in or deposit money, and compare it to other data points to determine whether individual customers might be on the verge of canceling their accounts.
  
• Smarter credit card and credit scoring: Determining creditworthiness is a critical banking service activity. Banks need to crunch significant amounts of customer data to make important credit decisions, such as whether they accept a credit card application or approve a credit increase. AI algorithms and machine learning can help financial institutions approve or deny credit cards, credit increases and other customer requests at fast speeds.

Introducing AI in banking is not without risks and complications. The IBM Institute for Business Value 2024 Global Outlook for Banking and Financial Markets study found over 60% of banking CEOs were concerned about new vulnerabilities introduced by AI. These include:

• Cybersecurity: Generative AI technology can be used for fraud prevention and compliance management, but it also produces risks. Embedding open AI tools and technologies into banking IT systems creates some security challenges because AI models are especially valuable targets for malicious actors. That’s why banks need a holistic AI governance approach that effectively balances innovation and risk management.
• Legal uncertainty related to operations: Generative AI models need training on existing data sets to be effective. There are still some unsolved issues on whether analyzing publicly available data, like news stories and explainer videos, constitutes copyright infringement⁴. One way to avoid this issue is to use AI models that have been trained on data that the bank owns, such as customer service interactions or its own proprietary research.
• Difficulties in controlling outcome accuracy: Currently, AI models do not reason or “understand” their outputs. Instead, AI models detect patterns⁵ in the data they’re given and generate results. Therefore, the model cannot tell the human employee if the data is incorrect or inaccurate.
• Prejudice from model bias: Banks are increasingly investing in environmental, social and governance (ESG) initiatives as a way to demonstrate transparency and accountability for their actions. Since AI models are trained on human-created data, they can inherit some of the biases that influence humans. Banks need to eliminate bias in how they market products and determine factors like creditworthiness, which historically has negatively affected certain demographics.

Banking institutions are under increased pressure for digital transformation. Customers demand automated experiences with self-service capabilities, but they also want interactions to feel personalized and uniquely human.

Banks continue to prioritize AI investment to stay ahead of the competition and offer customers increasingly sophisticated tools to manage their money and investments. Customers continue to prioritize banks that can offer personalized AI applications that help them gain visibility on their financial opportunities.

In the future, banks will advertise their use of AI and how they can deploy advancements faster than competitors. AI will help banks transition to new operating models, embrace digitization and smart automation, and achieve continued profitability in a new era of commercial and retail banking.