FL STEM Navigator: News From the Field
Audio Overview – a lively “deep dive” discussion that summarizes the key topics from this article
Morgan Stanley has developed an internal artificial intelligence (AI) tool, DevGen.AI, to address the persistent challenge of modernizing legacy code. Launched in January and built upon OpenAI’s GPT models, this tool translates outdated programming languages, such as Cobol, into clear English specifications, which developers can then use to rewrite code in contemporary languages.
Since its inception this year, DevGen.AI has processed nine million lines of code, resulting in an estimated 280,000 hours saved for Morgan Stanley’s development teams. Mike Pizzi, Morgan Stanley’s Global Head of Technology and Operations, noted that while commercial AI coding solutions are proficient in generating new code, they often lack the specialized knowledge required for older or proprietary programming languages. This gap prompted Morgan Stanley to create their own solution, allowing the tool to be specifically trained on the firm’s extensive and diverse codebase, including less common or custom languages.
The tool provides significant support to Morgan Stanley’s approximately 15,000 global developers. Its capabilities include translating legacy code into understandable English specifications, isolating specific code sections for regulatory inquiries, and performing partial translations of smaller legacy code segments into modern equivalents. Pizzi indicated that while the tool can technically rewrite code into new languages, its efficiency in producing fully optimized modern code is an area still under development, necessitating continued human oversight.
DevGen.AI’s most significant benefit lies in its ability to convert legacy code into clear English specifications, effectively creating a blueprint of the code’s function. This enables any developer to subsequently rewrite the old code in a modern programming language, mitigating the challenges posed by a dwindling pool of developers proficient in older or highly specialized coding languages. Morgan Stanley anticipates that AI will facilitate the development of more code and AI applications, thereby supporting business expansion and operational automation, all underpinned by a robust and standardized technological architecture.
