AI Expands Banks’ Options for Modernizing Legacy Cores

Artificial intelligence (AI) is providing banks with fresh avenues for modernizing their legacy systems and connecting them with newer applications. This advancement transcends chatbots and employee assistants. AI agents can extract data from banking systems, trigger software functions, transfer information across platforms, and manage specific workflow segments. Coding agents assist engineers in comprehending and modifying the software that links older and newer systems.

OpenAI's GPT-6 Astra, explored by PYMNTS, indicates that AI can manipulate existing applications through their interfaces, known as computer use. OpenAI also recognizes legacy-system modernization as a financial services application, including the migration of COBOL and other legacy code. Meanwhile, banking technology providers are making application programming interfaces (APIs) and core functionalities accessible to AI agents, offering additional methods to link newer applications with established systems.

A contemporary banking application generally interacts with a core via APIs, which retrieve account balances, create accounts, or carry out other predefined functions. Middleware between systems manages tasks like authentication, routing, and data translation. An AI agent can invoke these same APIs. Model Context Protocol (MCP) provides an additional means for the agent to discover available functions and how to utilize them, eliminating the need for developers to create separate AI integrations for each function.

Mambu's Core MCP serves as an illustration, exposing hundreds of core operations to AI clients. Specific functions that retrieve information can be made available independently from functions that modify information or execute actions, allowing institutions to maintain control over what an agent can perform. Computer use enables agents to navigate software using the same interface employees employ, granting access to applications that haven't been fully integrated through APIs.

This broad connectivity toolkit comprises APIs for direct system access, middleware for translation and routing, MCP for presenting approved capabilities to agents, and computer use for applications that heavily rely on screens and manual workflows. Banks are prepared to embrace more AI connectivity, as 95% of surveyed large financial services firms have broadly deployed or incorporated newer AI tools in data and technology processes.

However, integration still requires improvement. Only 25% of firms have adopted AI in API orchestration and integration. The leading obstacles to further AI deployment include data quality and fragmentation, with 30% of financial services firms identifying legacy technology as a hurdle to real-time payments modernization and 53% citing manual-intensive internal processes.

Addressing both the technology infrastructure and the associated workflows, connecting AI with existing systems tackles the technology estate and related processes. AI could also alter the economics for systems integrators like Accenture, Cognizant, and Capgemini. Coding agents can automate tasks related to mapping, coding, testing, and documentation involved in connecting bank systems, accelerating integration projects while providing firms with tools to expedite integrations.

MCP can simplify existing APIs for AI agents to discover and utilize. Nevertheless, as more agents are deployed, more connections must be governed, necessitating decisions from banks about system access, customer information retrieval, transaction initiation, and approval requirements for certain actions. Integration platforms can supply these controls, routing, monitoring, and audit records.

Ultimately, AI enhances what banks can offer users, from commercial lending applications that amalgamate customer and account data from multiple systems to servicing agents that compile payment histories and account records before an employee manages exceptions.