If anyone is interested, I wrote a blog that boils down recent research and studies of application dev on the mainframe using AI. You can read the article here on LI - https://www.linkedin.com/pulse/ai-assisted-application-modernization-mainframe-ai-doing-so8we/
Best regards, Tony Perri, CEO/Co-Founder [email protected]<mailto:[email protected]> www.linkedin.com/in/tonyperri/<http://www.linkedin.com/in/tonyperri/> ________________________________ From: IBM Mainframe Discussion List <[email protected]> on behalf of Brian Westerman <[email protected]> Sent: Tuesday, June 23, 2026 12:35 AM To: [email protected] <[email protected]> Subject: Re: AI on the mainframe I guess it depends on what you consider AI. We have been using AI on the mainframe in our automation software for over 10 years now, actually close to 20. The automation suite we sell has been making hierarchically based decisions based on previously noted information for quite a while. Does it think? No,, but neither does any of the other much newer A.I.'s. However, it does make decisions based on what it has learned from previous interactions, events and activities. It's a (not exactly quite) simple matter of keeping track of what and why you did something and providing a way to revisit that data and information and applying that knowledge at a later time when faced with the same or similar situation. You, as a person or operator, can interact with the automation suite and ask a limited subset of questions and allow it to make decisions or provide responses and answers based on your questions. The only real difference is that it's not really built to interact with real "physical" people outside of a data center environment, at least not very much, but we can email and text them information that we have been told that they are interested in knowing about when something happens. We have been telling our client sites for years that the automation suite can do almost anything that an operator can do except for the physical things like mount paper and change toner. We used to say we couldn't mount tapes, but now with virtual tape, we had to remove that as something we could not "do". :) The TELUM II CPU allows us a whole series of assembler instructions and paths that allow what we do internally to function better (i.e. much more efficiently). Most importantly it allows us to keep and use a MUCH larger pattern history table in memory, so not only can we make better informed choices, we can do them faster. So, you average internet user might not consider that "real" A.I. because they can't interact with it, but it works quite well, and we do it far cheaper than most, if not all z/OS automation suites. I just realized that this reads more like a marketing email, and that was not really my intent, so I apologize if it comes off that way. I really just meant to point out that A.I. has been around on the mainframe in a lot of ways for a very long time. It's more in how you use the features than if they have existed for a long time or not. Please don't get Dick started on how we should be afraid of A.I., it's not something that I want to deal with right now. :) Brian ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
