Where We Are And Where We Are Headed With AI On IBM i

February 16, 2026 Timothy Prickett Morgan

Here we are in 2026, and we have gotten used to GenAI like we did the coronavirus pandemic after it largely ran its course after three years. The difference, of course, is that there is no vaccine for GenAI, especially when a substantial part of the growth in the global economy is coming from enormous capital spending on GenAI hardware and software.

I have spent a large portion of my career as an IT journalist and analyst tracking high end, large scale, distributed computing in its many forms and have been fascinated by the clever and creative advances in the state of the art of compute, networking, and storage; sometimes, I have privately worried about the way it can be wielded for good or evil, but the job is to remain impartial and observe, so I don’t say too much in this regard.

What I can say is that I have the kind of skepticism and conservatism that I believe most OS/400 and IBM i folks espouse. And here, as the GenAI boom has gone from chemical to nuclear in its force, what I can honestly say is that I know for sure that I am for people. And that does not necessarily mean I am against GenAI, any more than I am not against nuclear weapons even though I think they are an extremely dangerous form of deterrence. I think we need to be thoughtful and careful about how the world does and does not use this GenAI technology, and what comes next.

So far, I am not all that impressed as the hyperscalers, cloud builders, and model builders are running, hell bent for leather, to a future economy they cannot and will not predict and one that I was not asked if I thought was a good idea after consideration. (Perhaps you were. I did not get the memo or the ballot.)

As GenAI matures and evolves, I see massive aggregations of money and power ahead, just as we saw in other infrastructure booms. But this one may be different in that with 1 billion AI agents, 1 billion AI-powered robots, and 1 billion autonomous vehicles, a lot of the work that the 3.3 billion people who are of working age on Earth currently get paid to do might be done by machines before too long. (There is another nearly 5 billion people on the planet who are children or who are old enough to be out of the workforce or lucky enough to be retired. They are not the same thing.)

No one in the political arena wants to stop the boom, Wall Street certainly doesn’t, the tech titans absolutely do not, and inasmuch as every retirement depends on the GenAI boom to sweeten it, many of us don’t, either. And as I said above, I get paid at my other job to analyze AI systems and the market for them. I watched this entire market emerge and evolve, in real-time. And I, too, am surprised by the emergent behaviors in neural networks that started out as language translators, and how this has changed into something more, however imperfect it may be at times.

It may turn out that some of the ways we store, retrieve, and synthesize data in our own brains is not so different from what a GenAI transformer model is doing when it statistically analyzes the connections between tokens of data in a massive dataset. As I write this, I have spent decades building my context window, which I can deploy for this very story you are reading, and I frankly have no idea what I am going to say next. Sometimes, I don’t know what I am thinking until my fingertips tell me. That is the God’s honest truth.

But, like you, I can make my neural network focus on a specific context and only chew on that data, something more specific about the IBM i base. In this case, I am referring to those parts of the IBM i Marketplace Survey put together by Fortra that speak to the adoption of AI into IBM i shops.

Artificial intelligence became a top five concern in the latest survey, which was completed last fall and compiled into a report in January for all of us to use – something that Fortra has supported and we have all benefitted from for the past dozen years. (This is exemplary of enlightened self-interest, for which we are grateful.)

The first time that Fortra added AI to its query about top concerns, where survey takers can pick their top five, was in the 2019 survey that resulted in the report published in January 2020 – which coincidentally was when the coronavirus was breaking our of Wuhan, China and when I came down with it after spending time in New York’s Chinatown shopping just ahead of the Christmas holiday in late December. (Yes, I do believe I am a superspreader, but no one knew about SARS-CoV-2 when I was in Chinatown or a few weeks later when I flew to San Jose to host an event for 250 people in the tech industry with my wife. I did not feel well at all before we left, and was told I had walking pneumonia, bronchitis, and pleurisy, given a Z-pack, and told to rest. Like I had time to do that. The next day I slept for 22 hours and the one after that, 18 hours. I felt more alive by day six.)

Here is how AI has moved up the top concerns rankings between 2020 and 2025, shown by share of respondents citing AI as a top concern:

2020: 12 percent
2021: 7 percent (odd that it got lower)
2022: 11 percent
2023: 10 percent (how slippery is this slope?)
2024: 18 percent (starting to climb)
2025: 30 percent

And here is where we are at, alongside all the other concerns, as AI was cited as a top concern by 42 percent of the IBM i shops surveyed for the 2026 report:

I fully expect for AI to keep moving to the left until it becomes the top concern, particularly because AI will become integral to solving the IBM i skills crisis, will be involved in cybersecurity, and will absolutely be a tool for application and database modernization. It will be hard to draw the lines between AI and not AI, in fact. What won’t be AI is perhaps a better question for all of us to ponder.

Last year, the IBM i Marketplace Survey took the pulse more accurately concerning the use of AI and hopes for the use of AI, and now that we have two years of data on these two questions, you can start to see some trends forming and why AI is moving up the top concerns rankings. I mean, aside from the obvious reasons that I cited at the top of this story.

The first is how companies that run IBM i systems for their back office (and possibly more) operations are currently using AI. It is important to note that this question is not restricted to AI running on IBM i systems, or even on premises, and just simply asks the IBM i survey takers how they are using AI. Take a look:

In one short year, the use of AI swung from 43 percent of those polled to 55 percent – that’s twelve points, exactly mirroring the twelve points of change of those who think AI is a top concern. The big jump came from using AI in application and other code development, which rose by 13 percent. We all know that code assistants are the killer app for IBM i shops when it comes to AI, given the skills shortage and the need to modernize applications. There is an increase in using AI for systems management and operations automation, and an even bigger jump for using AI as part of training. There is a slight uptick in AI being put into third party applications, but we are not sure if this is statistically significant given the size of the survey pool compared to the size of the IBM i base.

Now here is where it gets interesting. The next question has the same categories, but asks where IBM i could prove value for AI efforts in the future:

The future is not defined – two years from now, five years from now, and so forth – and that is intentionally so. If you put too many qualifiers on a question, you can’t get a gut reaction. This is about gut, the thing that people trust more than their own reason sometimes because it is often right about a lot of stuff.

As you can see, more than two thirds of those polled think code development is where IBM i and AI are going to provide value, and systems management and operations automation is a growing second choice. Ease of programming and automated systems are the hallmarks of the AS/400 design, so this is no surprise to me. People seem less concerned about data integration into AI for decision making (using retrieval-oriented generation, or RAG, techniques, presumably) of using AI to bolster security. For now, it looks like IBM i shops are looking for AI to help make it easier for companies that make and maintain their own code and run their own systems do more with less.

This is exactly what IBM’s strategy with integrated matrix math units on Power10 and Power11 processors and the Spyre AI accelerator cards for Power Systems and System z are all about. Keeping AI on the platform, and integrating it deeply and simply. And, we hope, cheaply. It would be hard to be more expensive than a Nvidia datacenter GPU these days. . . .

This Issue Sponsored By

Table of Contents

Content archive

Recent Posts

Subscribe

Pages

Search