Industry
Banking
Banking Services Transformation: Unlocking Exponential Value
Highlights
- Incumbent banks are under pressure to offer superior customer service and experience comparable with digital banks.
- Banks are hindered by inefficient back office operations characterized by multiple manual handoffs in meeting customer expectations.
- Enhancing customer service and delivering superior experience will require banks to move toward touchless operations by investing in technology.
On this page
Introduction
The banking and financial services industry is witnessing the rise of agile digital banks that offer innovative offerings and superior customer experience.
This is putting pressure on incumbent banks to increase operational efficiency and deliver comparable service. While banks have digitally transformed customer facing functions and interfaces, they are hindered by inefficient back office operations and struggle to meet customer expectations. Existing IT systems in banks are not yet fully capable of undertaking end-to-end straight-through processing (STP) of transactions without manual touchpoints. Despite significant advancements in technology, certain tasks remain unautomated, necessitating manual intervention from back office staff, hindering the shift to touchless operations.
If banks’ IT systems were advanced enough to seamlessly manage every step of the process without requiring manual intervention, the need for a back office would be significantly reduced or potentially eliminated. However, as it stands, the back office remains a critical component within banks. The reliance on manual processing has led to several challenges, including increase in transaction processing costs, longer turnaround time (TAT), and a higher likelihood of errors. The way forward lies in enhancing investment in emerging, cutting-edge technologies to streamline operations and reduce the dependence on back office functions. This will help banks to move toward touchless operations, raising the bar on customer service and experience.
Banks’ back office woes
Several factors hinder end-to-end automation of back office operations.
Manual intervention is often required for processing unstructured inputs, managing exception transactions, and making decisions that demand human judgement, experience, and knowledge. Tasks like data entry and hopping between multiple screens, document classification, and data extraction from various screens and applications are also manual.
Banks’ inability to achieve end-to-end back office automation, eliminate manual intervention, and enable touchless operations can be attributed to lack of integration of applications, gaps in functionality and process requirements, and the need to extract data from physical documents and images. For low-volume processes, the cost of automation often outweighs the benefits, making manual processing a more economical choice. Poor accuracy and the high cost and time required to train optical character recognition (OCR) systems for variations across document types make them less viable. Similarly, natural language processing (NLP) has been employed to handle unstructured inputs such as customer emails, call transcripts, and documents, but suffers from accuracy issues and high training costs. When it comes to handling exceptions and making decisions that are not based on a fixed set of rules, machine learning (ML) has shown some promise. However, it remains an expensive solution, limiting widespread adoption.
For banks, raising the bar on back office operations has become a business imperative, necessitating urgent action to overcome the impediments to making the back office smarter.
A blueprint for back office success
Automated or touchless processing is essential for delivering consistent customer experience and reducing the costs associated with banking operations.
We believe that banks must take targeted action across four key areas:
- Complex processes requiring manual intervention
- Processing of paper documents and images
- Processing of unstructured inputs
- Exception handling and investigation
In our view, back office transformation in banks will demand comprehensive adoption of artificial intelligence (AI) across the aforementioned areas.
Complex processes requiring manual intervention
Many processes in banks remain manual due to their complexity. Reliance on manual intervention highlights the need for more integrated and intelligent automation solutions. Bank staff often need to access customer information from a customer relationship management (CRM) system, verify transaction details in a separate financial application, perform rule-based processing, and then populate the output data in an accounting system. This multi-step, cross-application workflow is error-prone and prolonged, as it is dependent on manual data entry and validation. Automating such processes could significantly enhance efficiency and accuracy.
Figure 1: Process automation using AI agents
Low code, no code (LCNC) platforms and AI agents bolstered by large language models (LLMs) can integrate and automate business processes to enable STP quickly and cost-effectively (see Figure 1). With LCNC platforms, even non-technical users without technical expertise can design and deploy automated workflows, without the need for extensive coding. AI agents offer intelligent assistance, guiding users through the process of building natural language backed automation solutions, significantly reducing dependence on costly IT system upgrades and extended development cycles.
Processing of paper documents and images
In banks, physical documents are still prevalent across various lines of business (LoBs), creating significant challenges for efficient processing. For instance, loan applications, customer agreements, trade documents and account statements are often submitted in physical form. These documents are subsequently scanned and enter the system as images. Back office staff manually extract data from these documents and enter them into appropriate applications and systems, a time-consuming and error-prone task. A few institutions have leveraged optical character recognition (OCR) technology to automate this process but the results have been less than satisfactory, requiring. human intervention to correct errors and validate data, undermining the potential efficiency gains.
Figure 2: Data extraction using vision AI
Vision based LLMs with the capability to seamlessly extract relevant information (see Figure 2) from images can help. Unlike traditional OCR systems, vision-based LLMs can process and extract data with high precision from images without the need for prior training on specific formats. This will help banks eliminate manual processing and reduce costs and time, in turn enhancing operational efficiency, scalability, and adaptability.
Processing of unstructured inputs
Banks manually process unstructured data such as emails from customers, third parties, and internal systems, as well as communications from contact centers, collection departments, branches, and PDF documents. For example, a customer might send an email inquiring about the status of a loan application or a vendor might supply transaction details in PDF format. Banks are ill-equipped to manage the vast amounts of unstructured data flowing through these channels. Attempts to address this using NLP have been unsuccessful due to high costs and lack of accuracy and scalability.
Figure 3: Processing of unstructured data using AI agents
LLMs can help address the limitations of banks in handling unstructured data by accurately interpreting and processing it (see Figure 3). Unlike traditional NLP systems, LLMs can understand context, extract relevant information, and generate structured outputs without extensive training or manual intervention allowing AI agents to process transactions straight-through. For example, suppose a customer sends an email enquiring about the latest interest rates on personal loans, an AI agent can read the mail and send a personalized response relieving back office staff of such mundane tasks. This capability not only reduces the cost and complexity associated with processing unstructured data but also enhances scalability and reduces turnaround time.
Exception handling and investigation
Exception processing and investigation are critical areas that often require manual intervention due to the absence of fixed rules. For example, in fraud detection, when a transaction is flagged as suspicious, it typically requires a detailed investigation by a fraud analyst who examines various data points, such as transaction history, customer behaviour, and external factors, to determine whether the transaction is indeed fraudulent. This process relies heavily on the analyst's experience and intuition, making it difficult to automate. Similarly, in dispute management, when a customer objects to a charge, the case must be thoroughly investigated to gather evidence, communicate with the parties involved, and make a fair decision. This involves analyzing transaction records, customer communications, and sometimes even legal documents, all of which require a nuanced understanding that existing IT systems struggle to replicate.
AI agents built using LLMs with reasoning capability, offer a promising solution to the challenges of exception processing and investigation. For instance, in fraud detection, AI agents can continuously learn from new data and past frauds to improve the ability to detect suspicious transactions and recommend next best actions. They can analyze transaction histories, customer behaviours, and external data sources in real-time, eliminating the need for manual analysis. Similarly, in dispute management, AI agents can automate the gathering and analysis of evidence, streamline communication with the parties involved, and suggest fair resolutions based on historical data and contextual understanding (see Figure 4).
Figure 4: Exception handing using AI agents
In trade finance, where transactions often involve multiple parties, complex documentation, and cross-border regulations, AI agents can significantly enhance efficiency. For example, when discrepancies arise in letters of credit or shipping documents, AI agents can quickly analyze the documents, identify inconsistencies, and suggest corrective action. They can also assist in compliance checks by cross-referencing transaction details with regulatory requirements, reducing the risk of errors, and ensuring timely processing. By automating these intricate tasks using AI agents, bank staff can focus on more complex and nuanced cases, ultimately improving overall operational effectiveness and customer satisfaction. Autonomous AI agents can significantly increase STP. However, high value transactions or those necessitating manual verification due to regulatory requirements will be transferred to human agents for final review and approval.
According to the TCS AI for Business Study, 43% of banking, financial services, and insurance (BFSI) firms expect more than half their employees to be using generative artificial intelligence (GenAI) on a daily basis within the next three years. In our view, in the foreseeable future, AI agents will coexist with human agents in banks’ back office. Essentially, banks can rely on a vast army of ‘digital’ agents, assisting in the swift execution of complex, time-consuming, and knowledge-intensive tasks, thereby dramatically enhancing the overall efficiency and productivity of the back office.
Looking ahead
LLMs and AI agents offer immense potential to transform back office operations in the financial services industry.
Reinventing the back office operating model by harmoniously integrating AI with human effort will free banks’ staff of routine, monotonous work, allowing them to take on higher value-adding tasks and positively contribute to the financial institution’s performance.
An efficient back office means superior customer experience and delight. With the commoditization of the financial services industry, experience is a key differentiator. Given the heightened importance customers from the millennial and GenZ segments place on experience, banks must act quickly to transform their back office and gain a lead in the competitive financial services industry.
