- The newly patented DevGen.AI tool was developed in-house and released in January.
- The bank says it's saved almost 300,000 hours this year alone, but won't replace developers' jobs.
- DevGen.AI solves a long-standing headache for coders within and beyond Morgan Stanley.
During a Morgan Stanley hackathon at the end of 2023, two teams were trying to solve a problem that has plagued developers for years: how to rewrite legacy code into modern programming languages.
Those nuggets of an idea turned into DevGen.AI, a tool the bank unveiled in January that has so far saved developers more than 280,000 hours as of June, or 11,666 days they would have previously dedicated to deciphering outdated code, according to Trevor Brosnan, the bank's global head of technology strategy, architecture, and modernization. Brosnan led the effort to build the tool, which turns code from old languages into plain-English specs that can then be rewritten.
Legacy code refers to outdated, older software code — think of languages like Cobol, which was developed in 1959 — that can raise security risks and slow down how quickly companies can take advantage of new technology, The Wall Street Journal reported. Banks and financial institutions are especially dependent on older programming languages, per the Journal.
"In early 2024, I pulled some of our top distinguished engineers at the firm out of their line of business teams and into a new applied AI team, because my instincts told me that there was some great opportunity here," Brosnan told BI.
His instincts were right — the tool has exceeded even his high expectations. It was granted a patent in early June with assistance from the bank's Patent Accelerator Program, which Megan Brewer, the head of firmwide market innovation and labs, helped start.
Cutting out 'onerous' work, not jobs
Brewer and Brosnan said that DevGen.AI won't take software engineering or other jobs with the time it saves. The technology, Brewer said, replaces "onerous" rote work.
"That means those people can actually go work on what we need to be working on, which is the future," she said. Brewer said AI presents opportunities throughout the firm, and that Morgan Stanley is currently looking into how it can use agentic AI across teams.
"The reality is we have so much modernization work to do, and we have ongoing demand from all of our businesses to deliver more functionality, more capability for our clients," Brosnan told BI.
Young people are struggling to land quality positions in the tech sector, in particular, and AI is already starting to gobble up jobs across industries. White-collar job postings nationwide are shrinking faster than blue-collar listings, including for software developers.
Still, Brewer and Brosnan were adamant that DevGen.AI won't render developers obsolete. Morgan Stanley had 233 technology jobs in the Americas posted on its website at the time of writing, including dozens of openings for software engineers.
The initial team for DevGen.AI was tiny
Not many of the technologists at the bank worked on DevGen.AI — Brosnan said the initial team was fewer than five people "who are passionate about this." Eventually, the core group expanded to around 20 engineers. Throughout the process, they consulted with subject matter experts across the firm to figure out different use cases for the tool, Brosnan said.
Morgan Stanley could have outsourced the development of the tool, but Brosnan said they decided to develop it internally in part to "implement all the kinds of security controls." Now, he said, employees across divisions, from its institutional to wealth management businesses, are using the tool.
"Even within the world of generative AI, this was a very, very, very clever use of it, but it also was extremely impactful," said Larry Bromberg, the global head of intellectual property legal, who co-leads the Patent Acceleration Program.
Pleased as he is with the result, Brosnan said his team very briefly celebrated their successes.
"Frankly, they are still very focused on their day job. Our mission is to accelerate modernization at Morgan Stanley, and we've still got lots of work to do," he told BI.
The technology will stay internal for now
One cause for celebration was receiving a patent, which Morgan Stanley employees are encouraged to do through the Patent Accelerator Program. Brewer said the program provides support throughout the invention and legal process.
Patent-holders span the firm — there are around 500 across 14 different divisions, Brewer said. From 2023-2024, the Patent Accelerator Program increased final submissions to patent offices by 53%, and Brewer said that they'll "definitely" surpass that this year. Brewer said that 20% of patent holders are junior staff, and 10% are distinguished engineers.
One of the lead engineers on the DevGen.AI team is something of a serial patent recipient, or a "serial offender" as Brewer put it, with 13 patents to date.
Brosnan said the team doesn't have immediate plans to license the newly patented tool externally; it would likely be in high demand given how many financial and Big Tech companies have to deal with outdated coding languages.
"Right now, our plan is to leverage this definitely for all of our modernization demands internally. We haven't decided anything different from that," he said, adding that Morgan Stanley has shared technology with peers in the past.
In the meantime, the question of how to use AI and modernize the bank isn't going anywhere. As Brosnan put it, his technologists are back to "hands-on keyboards" after every win.
