Why Mainframe Exit Projects Fail in the Age of GenAI: An Ensono Q&A

The Gartner report predicts that more than 70% of mainframe exit projects initiated in 2026 will fail to deliver intended benefits, largely due to overestimation of generative AI capabilities, and that 75% of vendors in the mainframe exit market will pivot or cease to exist by 2030.

The recent Gartner® report, “Too Big to Fail: Why Mainframe Exit Projects Are Likely to Fail in the Age of Generative AI,” predicts more than 70% of mainframe exit projects initiated in 2026 will fail to produce the intended benefits due to an overestimation of generative AI tooling capabilities and by 2030, 75% of vendors operating in the “mainframe exit” market will either pivot their business models or cease to exist.

We sat down with Lisa Dyer, SVP of Mainframe Line of Business, and Brian Klingbeil, EVP and Chief Strategy Officer, to discuss what this research means for enterprise leaders—and Ensono’s perspective on the path forward.

What We Think This Report Signals

Q: What does this Gartner report indicate about where the market is headed?

Lisa: It witnesses a pivot—a confluence of things that have happened in the market that cause organizations to take a step back and reevaluate the place of their mainframe in their overall strategy. Organizations have seen the fallout from setting their strategy to simply get off the mainframe. Those projects have largely not panned out in the way expected—overruns in cost, the end state not behaving the way it should. It may be harder than they thought. And now with the evolution of AI possibilities on the platform, many organizations are revisiting the strategy to validate or pivot.

Brian: This report is an interesting catalyst for people to get a deeper understanding of what the strengths and weaknesses of the platform actually are. I’ve said this before, and I’ll keep saying it: The emergence of all these AI tools is actually making it easier to stay on the platform, not easier to leave.

For 10 years, there’s been a propensity for people to conflate mainframe modernization with mainframe migration. Modernization is a higher-level principle. Sometimes migration is one element of modernization. What the industry has found over the last decade is that some applications are well-suited for the mainframe, and some aren’t. The mainframe’s strengths are massive transaction processing at scale, high uptime that’s difficult or impossible to replicate in cloud, and security that’s equally hard to match. The weaknesses have been agility, difficulty understanding legacy code, and getting support from a maturing workforce. What’s interesting is that AI tools are now making it easier to remediate those weaknesses. It’s all coming to a head, and this report does a great job of summarizing that.

Why Mainframe Exit Projects Fail

Q: Gartner predicts more than 70% of mainframe exit projects initiated in 2026 will fail to deliver intended benefits. What, according to us, is driving that disconnect?

Lisa: One reason for the disconnect is the misconception that the problem you’re trying to solve— a lack of agility, velocity, and skills to continue with your applications running on mainframes—is only solved by moving to cloud. Conflating the need to modernize with the need to migrate is common. It’s rarely, if ever, that binary.

Brian: That prediction isn’t a stretch because that’s pretty much what’s been happening. If people don’t understand why the mainframe is great at what it does, they’ll keep making the same mistakes. This report is a great catalyst to get people to realize that certain applications are extremely well-suited for the mainframe. Previously, people looked at it as an expensive platform and assumed that if they could leave, they should. But it’s not a question of whether you have the ability to leave. It’s whether it even makes sense.

The GenAI Reality Check

Q: Gartner points to overestimation of generative AI tooling capabilities as a key driver of these failures. Where do you see GenAI adding real value—and where is the market overpromising?

Lisa: When the conversation focuses only on code—converting COBOL to Java, for example—that’s the tip of the iceberg. Underneath that is the system itself: how it’s architected to run that code optimally, the transactional throughput, the performance characteristics. All of that is baked deep into the platform. The promise that you can point an LLM at your code and move everything to the cloud misses the wholesome systems thinking.

Where GenAI does add tremendous value is in lowering the barrier to operating and evolving the mainframe apps and code jobs to be done. As LLMs get better at understanding mainframe environments—the infrastructure, the applications, the context—they help teams understand what they have, develop faster, and continuously improve. That’s real value. But an easy button to migrate everything wholesale? I don’t see that.

Brian: The people selling the GenAI exit dream are probably the ones who stand to bill a lot of hours using those tools—without delivering the outcome. The smarter players are recognizing that these tools often strengthen the case for staying on the mainframe.

It’s now easier to understand your legacy code, reduce dependency on an aging workforce, and develop with modern tools and processes. 18 months ago, a lot of this wasn’t even possible. A whole new world of opportunity has opened up to make the mainframe a more agile and innovative member of your IT ecosystem—and enterprises are just now learning this is the case.

Buying Direct Versus Working With a Partner

Q: What’s our guidance on how clients should acquire software?

Lisa: I’d say organizations should understand their options. Dominant MSPs like Ensono have buying power that most organizations simply don’t have on their own—that’s a fact. Beyond cost, MSPs help clients actually leverage the technology because we’re experts in it. I recently spoke with an organization that had upgraded their mainframe database but had no idea what new capabilities were available or how to use them. They needed our experts to help them understand what’s possible. That’s value you don’t get from a purchasing agreement alone. But again—not a binary choice. Between Ensono and IBM—getting the best of what both have to offer—that’s a true force multiplier.

Brian: It depends on the client and their size. But as a service provider, IBM is our most important partner—we’ve done massive business with them for 10 years and will for the next 100. And because of that scale, we have purchasing power and operational advantages that individual organizations can’t get on their own—the ability to amalgamate traffic across clients, share storage and infrastructure, trade in old hardware, and recycle components. If you have a complex setup, you wouldn’t want your own underutilized infrastructure.

That said, I’d encourage clients to have a relationship with both Ensono and IBM. If you’re our client, you’re also IBM’s customer—and everybody wants to keep their customer happy. You can lean on two providers instead of one, and a lot of smart people are at your disposal.

The Vendor Shakeout

Q: Gartner predicts that by 2030, 75% of vendors operating in the “mainframe exit” market will either pivot their business models or cease to exist. What, according to us, does that say about the promises being made today?

Lisa: It’s a bold statement, but it’s not ridiculous. The key word is “pivot”—and that’s a flexible term. LLMs are democratizing the landscape, just like they’re disrupting the SaaS business. Vendors who’ve built IP over the years without that democratization are going to have to adapt. The ones who find the right niche, pivot, and focus on in-place modernization of mainframe applications—those will be the winners. The ones who don’t will be going out of business.

Brian: I agree with Lisa on the word “pivot.” When organizations realize that not only is it hard to move off the mainframe, but they actually don’t want to—because it’s quite good at what it does—those vendors should shift toward helping clients leverage AI to take advantage of what’s now possible on the mainframe, rather than helping them leave it.

Mainframe Misconceptions

Q: What’s the biggest misconception enterprise leaders still have about their mainframe?

Lisa: There are a lot. Myth one: Mainframes can’t use AI. Myth two: Younger generations don’t have the skills to innovate on them. Myth three: Nobody interacts with the mainframe. Myth four: You can’t access mainframe data from other platforms. Myth five: It’s too expensive. Too expensive relative to what? The TCO matters relative to what the platform is doing for your business and your revenue-driving apps that run on it.

Brian: We see younger CIOs stepping into well-established companies with mainframes, and sometimes they’ve never seen one before. They assume it’s a relic that should have died in the 80s. Not only did it not die, it’s been constantly modernized, and it still runs the vast majority of transactions in the world. We joke that if you turned off all the mainframes tomorrow, there would be a zombie apocalypse. IBM estimates that roughly 78% of the world’s transactions still go through mainframe. So, you get a CIO who assumes it’s old and should be replaced, without stepping back to realize it’s not a big black scary box. It’s a collection of applications, each with its own characteristics. You have to evaluate based on that merit.

The Path Forward

Q: If a full mainframe exit isn’t viable for most enterprises, what should leaders be prioritizing?

Lisa: Here’s part of the problem: You ask 10 different people what mainframe modernization means, you get 12 different answers. People conflate platform modernization with application modernization, and then within application modernization, there are multiple paths—optimizing in place, changing languages while staying on the platform, or actually moving workloads elsewhere. Those distinctions get lost all the time.

Organizations should think about what modernization means for you. Why is it important? What problems are you trying to solve? What competitive imperatives is your business talking about in your 10-K? Answer those questions first, then back into the technology. Starting with the tech is why many of those projects failed. It was mainframe to cloud, for example-that’s a technology conversation, not a business one.

De-risk. Optimize. Get rid of some of the technical debt you have. Tune what you have to perform better and create cost efficiencies. Reset the foundation from where you can innovate and be disruptive in your industry category. And if you haven’t yet, get objective expert insights about what you can do with latest generation of the platform.

Brian: Maybe don’t start with worry about if you can exit the platform, but start by asking: Do I even want to? Is this the best execution venue for what I’m doing? If the answer is yes—it’s great at what it does, can’t be replicated in cloud, or it’s too risky and expensive to move—then turn your attention to what’s now possible. AI augmentation of the mainframe wasn’t possible 18 months ago. Now it is. You can get 95% of the benefits you’re looking for with a fraction of the cost and a fraction of the risk.

The approach should be application disposition. Look at everything that runs on the mainframe, assess its business-critical nature, and ask how it could be made better. A lot of leaders are running around with an AI hammer looking for nails. Flip that equation—start with the business outcome you’re looking for, then decide what tools are right for the job.

---

Frequently Asked Questions

Why do mainframe exit projects fail?

According to Gartner, more than 70% of mainframe exit projects initiated in 2026 will fail to produce the intended benefits due to an overestimation of generative AI tooling capabilities. Organizations often conflate the need to modernize with the need to migrate, underestimating the complexity of replicating mainframe performance, security, and reliability in cloud environments.

What’s the difference between mainframe modernization and migration?

Modernization is a higher-level principle that includes optimizing applications in place, updating languages while staying on the platform, or moving select workloads elsewhere. Migration refers specifically to moving applications off the mainframe. Migration may be one element of modernization—but it’s a small element, and not always the right choice.

Can generative AI help modernize mainframes?

Yes. GenAI tools can help teams understand legacy code, reduce dependency on aging workforces, and accelerate development using modern tools and processes. However, GenAI isn’t an easy button for wholesale migration—it’s better suited to in-place modernization and operational improvements.

Should enterprises work with an MSP or buy mainframe software directly from IBM?

It depends on size and needs. MSPs typically offer greater buying power, shared infrastructure economics, and expertise in leveraging the technology—benefits most organizations can’t achieve on their own. Enterprises can also maintain direct relationships with IBM while working with an MSP.

Gartner, Too Big to Fail: Why Mainframe Exit Projects Are Likely to Fail in the Age of Generative AI, By Dennis Smith, Alessandro Galimberti, Tobi Bet, 8 April 2026. GARTNER is a trademark of Gartner, Inc. and/or its affiliates.