Building efficient banking operations with generative AI

Arun Menon,

Head - Intelligent Process Automation, BFS North America

Published: June 10, 2024

It’s difficult to ignore the disruptive potential of generative artificial intelligence (genAI). While users typically praise genAI’s creativity and its reasoning abilities, genAI tools can also improve operational efficiency and increase cost savings. Popular applications of genAI include creative uses: writing articles, designing marketing materials, and generating art. However, because the best genAI models can engage in humanlike reasoning, understand and generate natural language, process images, and even simulate human interactions, these tools hold enormous potential for streamlining operations, slashing costs, and boosting overall business efficiency.

This news is particularly important to the banking and financial services (BFS) industry. Despite the increasing popularity of automation technologies, the banking industry has long prioritized highly manual workflows; multiple and disparate structured and unstructured data sources; and legacy systems. Generative AI can unlock significant efficiency gains and cost reductions across multiple banking verticals and across the customer life cycle, from customer onboarding and transaction processing to risk management and customer service.

Below are some ways in which genAI can streamline workflows across various lines of business (LOBs) in banking.

Retail banking

  • Mortgage processing:  Generative AI models can automate the extraction of data from loan applications, income statements, and identity documents to generate personalized loan application forms with prefilled fields. In doing so, these models can streamline the onboarding process, accelerate loan approvals, and ultimately provide a more efficient customer experience (CX).
  • Customer service: By processing data from historical customer chat and call logs with large language models (LLMs), agents can anticipate and handle customer needs more effectively. By identifying common queries, inquiry trends, and complaint patterns, agents can provide proactive assistance, thus enhancing customer satisfaction scores (CSAT) and reducing time to resolution (TTR).

Commercial banking

  • Covenant management: LLMs can analyze commercial loan contracts to automatically identify and classify covenants, both financial and non-financial. By generating summaries of complex loan documents, these tools enable loan officers to quickly grasp key covenant details and potential risks, improving decision-making and risk management processes.
  • Trade finance: Multimodal LLMs can automate classification and data extraction from trade finance documents, such as letters of credit, guarantees, bills of lading, and invoices. This process streamlines import and export payments, reduces manual effort, and facilitates smoother transactions and shorter processing times.

Investment banking and capital markets

  • Investment research: LLMs enable rapid analysis, summarization, and generation of insights by processing market trends, financial statements, security filings, and news. By doing so, the LLMs can assist portfolio managers, wealth advisors, and traders in building and recommending investment strategies more efficiently.
  • Pitchbook preparations: By automating the pitchbook creation process, generative AI technology can analyze relevant financial data, generate visualizations, and compile information into cohesive presentations for investment banking pitches. Additionally, the technology can create content for investment banking reports by compiling information from multiple data sources.
  • Trade settlement: Multimodal LLMs can automate the extraction and analysis of trade data in trade settlement emails and messages, thus accelerating trade detail verification and streamlining trade settlement operations.

Risk and compliance

  • Anti-money laundering (AML) research: Generative AI models can process information from multiple data sources, including free and paid sources (e.g, LexisNexis), to prepare draft suspicious activity reports (SARs) on individuals and entities undergoing investigation, thus reducing compliance costs and improving AML workflow efficiency.
  • Regulatory reporting: Large language and vision models can accelerate regulatory filing preparation by analyzing and compiling data from multiple structured and unstructured data sources, thus helping financial institutions efficiently meet regulatory requirements and reducing compliance-related risks.


  • Invoice processing: Multimodal LLMs can extract data from invoices and reconcile  purchase order data with invoice fields, thus streamlining accounts payable operations, accelerating payment processing, and improving overall efficiency.

While genAI holds immense potential for transforming banking operations, it is critical that users recognize the importance of human expertise in vetting AI outcomes, particularly in the context of the industry’s strict compliance requirements. Human oversight and continuous monitoring remain essential to managing the risks native to genAI technology, (e.g., hallucinations, poor explainability, and training data bias). Additionally, collaborations between data scientists and finance professionals will further ensure proper AI governance.

As genAI tools evolve, banking leaders and customers will encounter a growing number of applications and  risks. While genAI is poised to transform the banking industry landscape, proper governance is key to unlocking the technology’s potential. 

Generative AI 

Illuminating the next generation of enterprise innovation

Related content