Question for those still using Cobol, do you trust GenAI-based coding assistants to translate Cobol correctly?
Yes, limited need for Cobol knowledge24%
Somewhat, but still need significant Cobol knowledge56%
No15%
Not sure yet4%
94 PARTICIPANTS
Yes, limited need for Cobol knowledge24%
Somewhat, but still need significant Cobol knowledge56%
No15%
Not sure yet4%
More invested in GenAI33%
Investment is evenly distributed48%
More invested in legacy AI16%
Not sure1%
Mainframes will be replaced with cloud based resources in 10 years or less74%
Mainframes will continue to survive for at least 10 more years.25%
No selling.
No recruiting.
No self promotion.
Rules of EngagementFAQsPrivacy
© 2025 Gartner, Inc. and/or its affiliates. All rights reserved.
Over the last 2-3 years, we look to several solutions that would allow to REFACTOR Cobol applications to Java, for instance.
Then, we realized that our legacy system (built using Cobol, IMS FastPath Transactions, a bit of Assembler here and there) was quite complex.
Then, we looked at generating Cobol Code (since our challenge was to maintain sufficient number of developers) as it would relief the pressure to find more people.
But the code had to 'fit' in place with a very strict 'framework' we built. The result wasn't absolutely satisfying.
Our legacy system has around 15 millions lines of code and uses a lot of common modules.