AI Takes on Legacy Code: ChatGPT and Claude Could Revive Ageing COBOL Systems
Poulami Saha
COBOL’s Six-Decade Legacy
Created in 1959, COBOL continues powering critical banking and government systems, despite its ageing infrastructure and shrinking developer community.
The Growing Programmer Shortage
Experienced COBOL programmers are retiring, creating maintenance challenges for organisations dependent on decades-old mainframe software and specialised technical knowledge.
OpenAI’s AI-Powered Modernisation Push
OpenAI believes advanced ChatGPT models can help migrate ageing COBOL applications into modern programming languages while preserving their original functionality.
Banking’s Dependence on COBOL
According to IBM, COBOL supports over 40 per cent of online banking systems, 80 per cent of in-person credit card transactions and 95 per cent of ATM transactions.
GPT-6 Astra Shows Improved Performance
GPT-6 Astra achieves a 72.1 per cent COBOL test pass rate, significantly improving upon GPT-5.6 Sol’s earlier 26.8 per cent performance.
Multiple Modern Programming Languages Supported
OpenAI says AI models can translate COBOL applications into C++, Java, Kotlin, Python or Rust while maintaining their existing operational behaviour.
Anthropic Enters the Modernisation Race
Anthropic has also positioned Claude Code as a potential solution for modernising COBOL applications running on banking, insurance and government mainframes.
Legacy Software Creates Economic Challenges
A 2024 Brookings report estimated COBOL inefficiencies cost the United States approximately $105 billion annually, highlighting modernisation's economic importance.
Human Oversight Remains Essential
AI-powered migration remains incomplete, requiring experienced developers to review converted applications, validate business rules and conduct extensive production testing.