AI in commercial banking: Transforming payment processing and beyond

Commercial banks have long struggled to balance cost optimization with revenue growth, a challenge particularly evident in payment routing.

Traditional systems rely on rigid, predetermined processes that often lack adaptability in today’s fast-paced financial environment. These outdated methods are inefficient, overlook opportunities for cost savings and revenue enhancement, and create silos that hinder collaboration across departments and regions. This fragmentation slows decision-making, increases operational costs and results in suboptimal client experience.

Enter generative and agentic AI: Catalysts for transformation

Most payment routing today is based on rule-based strategies which are broadly applied across thousands of transactions and not contextualized for each individual transaction. This process is prone to delays, errors and inefficiencies, limiting the ability to optimize success rates for transactions. Payment failures are often unpredictable, resulting in frustration for merchants and customers alike.

Generative artificial intelligence (gen AI) and agentic AI present a paradigm shift in how banks can approach payment routing and related processes. Gen AI excels at analyzing historical data from vast amounts of structured and unstructured sources (like audio, images, text) and predicting optimal routes for payments. Agentic AI takes it a step further by acting autonomously within predefined guidelines to make decisions and execute actions in real-time, reducing the need for manual intervention and accelerating processing times.

For example, gen AI can analyze logs and system data to identify vulnerabilities, and agentic AI can dynamically execute payments via more secure pathways. Both systems can work in tandem to identify the most cost-effective routes for transactions, considering factors like currency exchange rates, processing fees and time zones, dynamically adjusting to changing market conditions, ensuring that payment decisions remain relevant and timely. They break down silos by intelligently integrating data from multiple internal and external sources, systems and departments, enabling holistic analysis and the creation of a unified view of the entire landscape.

Through a context-aware, multi-stage agentic AI powered process, transactions can be routed dynamically and in real-time. These are:

1. Analysis: When a transaction is initiated, associated data (such as card issuer, transaction value, and time of day) is analyzed in real time to calculate acceptance probabilities and route costs across available acquirers.
2. Optimization: Based on this analysis, the system identifies the optimal acquirer or pathway, ensuring the highest success rates while minimizing costs. This happens in milliseconds, ensuring both speed and accuracy.
3. Continuous Learning: AI systems refine their decision-making processes over time by learning from past successes and failures, dynamically adapting to acquirer behavior, transaction patterns, and technical issues to further optimize routing strategies.

Beyond payment routing

The integration of generative and agentic AI with traditional banking capabilities enables banks to deliver significant value across key areas:

1. Client Liquidity Management: Generative AI can forecast clients’ liquidity needs with high accuracy by analyzing historical transaction data and external market trends. It offers personalized recommendations for optimal cash positioning and investment strategies, helping clients maximize returns on their idle funds while maintaining necessary liquidity buffers. Agentic AI can autonomously recommend or implement liquidity solutions in real-time, ensuring clients have access to funds when needed.
2. Cash Flow Analysis: AI models predict cash flow patterns and assist clients in managing working capital. Agentic AI proactively identifies potential cash flow issues and suggests and takes corrective actions.
3. Creditworthiness and Stress Testing: Traditional credit assessment models often struggle to capture the full complexity of market dynamics and their impact on client creditworthiness. Gen AI and agentic AI can perform continuous stress testing by collating information from multiple risk intelligence sources in real time and autonomously flag or mitigate risks, providing banks with a more resilient credit portfolio.

The road ahead

Adopting gen AI and agentic AI comes with challenges, including data privacy, security, regulatory compliance, change management and integrating AI with legacy systems. However, these hurdles can be managed with robust strategies and strong partnerships.

By transforming static processes like payment routing into dynamic, adaptable and intelligent systems, these technologies drive efficiencies and open new growth opportunities. Banks that embed gen AI and agentic AI into their core operations will unlock innovative growth engines and remain competitive in an evolving landscape.

Vivek Dwivedi is U.S. Regional Head, Payments and Cards at Infosys.