Morgan Stanley’s AI Read 9 Million Lines of Code Nobody Understood – And Saved 280,000 Developer Hours
9 million lines of legacy code. Written decades ago in COBOL. Still running systems that move billions of dollars daily. And until recently, nobody alive fully understood all of it.
Morgan Stanley just showed what happens when you point AI at one of the oldest, most expensive problems in enterprise technology – and solve it.
The Problem Nobody Talks About
Here’s something most people outside of tech don’t realize: the biggest banks, insurance companies, and government agencies in the world still run on code written in the 1960s and 70s. Languages like COBOL – older than the internet itself.
The original developers are retired or gone. Documentation is sparse or missing. And the code is so deeply embedded in critical systems that changing one line could crash operations that handle billions in daily transactions. So companies don’t touch it. They build workarounds. They layer new systems on top. And the technical debt compounds year after year.
What Morgan Stanley Built
In January, Morgan Stanley rolled out DevGen.AI – an AI tool built in-house on OpenAI’s GPT models, trained on the company’s own proprietary codebase.
DevGen.AI doesn’t replace code automatically. It does something arguably more valuable: it reads legacy code line by line and translates it into plain English specifications. For the first time, modern developers can understand what decades-old systems actually do – and rewrite them safely in modern languages.
The Numbers
The results are staggering. DevGen.AI has processed 9 million lines of legacy code. It’s saved approximately 280,000 developer hours. And 15,000 developers have shifted from manual code translation to strategic product work.
The tool processes code at about 32 lines per hour – which sounds slow until you realize the alternative was developers spending weeks manually reading code that nobody alive wrote, trying to figure out what it does before they could change a single line.
Why This Actually Worked
The critical insight is that Morgan Stanley didn’t use an off-the-shelf AI coding tool. Commercial tools are excellent at writing new code but lack expertise in company-specific legacy systems and older languages like COBOL.
By training the AI on their own codebase, Morgan Stanley created a tool that understands their specific systems – something no general-purpose AI could do. Domain-specific AI beat general-purpose AI decisively.
I’m Mike Partners – entrepreneur, investor, and founder of AiExpert.org. I write these breakdowns because every business deserves access to the strategies that are reshaping entire industries. Here’s how to act on this one.
How to Apply This to Your Business
Even if you’re not a tech company, you probably have processes that could benefit from the same automation-first thinking. If you do have developers or technical staff, introduce AI coding assistants and measure the productivity gain over 30 days. If you don’t, think about what manual processes in your business could be replaced by simple software tools – many of which can now be built with AI assistance even by non-technical founders. The barrier to custom business software has never been lower.
The SMB Playbook
1. Identify your “legacy code.” You might not have COBOL, but you almost certainly have old spreadsheets nobody fully understands, processes that only one person knows, or custom software built by someone who left years ago. That’s your technical debt.
2. Use AI to document before you rebuild. The biggest lesson from Morgan Stanley is: understand first, change second. Use tools like Claude, Cursor, or ChatGPT to read and explain your existing systems before you try to replace them.
3. Start with the system that scares you most. The one everybody’s afraid to touch? That’s the one creating the most hidden risk and the most opportunity. AI documentation turns the unknown into the manageable.
Frequently Asked Questions
How is Morgan Stanley using AI in 2026?
Morgan Stanley has deployed AI across multiple areas of its operations, focusing on automation, cost reduction, and efficiency gains. As covered in this analysis by Mike Partners, the results include measurable improvements in both operational metrics and financial performance, demonstrating that strategic AI deployment delivers real business returns.
What business results has Morgan Stanley achieved with AI?
Morgan Stanley has demonstrated that AI can drive meaningful improvements in both efficiency and financial performance. The key results include reduced operational costs, improved productivity per employee, and faster execution on core business processes.
How can small businesses apply the same AI strategies as Morgan Stanley?
Small businesses can apply similar principles by starting with their most repetitive, time-consuming processes and finding affordable AI tools to automate them. Resources like AiExpert.org break down enterprise AI strategies into actionable steps sized for smaller companies, so you do not need a Fortune 500 budget to benefit from these approaches.
What is the ROI of AI automation for businesses in 2026?
ROI varies by implementation, but the pattern across major deployments is consistent: companies are seeing 20-40% cost reductions in automated processes, significant productivity improvements per employee, and faster decision-making cycles. The key driver of ROI is not the technology itself but how strategically it is deployed against the business’s highest-cost, most repetitive operations.
What AI tools should I use to automate my business like Morgan Stanley?
The right tools depend on your specific business needs. For customer-facing automation, look at chatbot platforms and AI-powered support tools. For operations, explore workflow automation platforms like Zapier or Make. For content and marketing, tools like ChatGPT, Jasper, or Claude can accelerate production. Start with one area, measure results over 30 days, and expand from there.
