Deluxe Corporation beats the odds with mainframe migration using AI

Deluxe may have prevailed against the odds when it successfully migrated from a 50-plus-year-old mainframe recently.

The company was able to move from it to a cloud environment in about 12 months without a major hitch, and while AI did some of the heavy lifting. The save will amount to about $4.9 million a year by retiring old hardware and software, cutting labor costs, and consolidating IT resources.

Deluxe, based in Minneapolis and traditionally known as a check printer, has transformed itself into a payments IT provider in recent years. But it desperately needed to end its reliance on its ancient mainframe, says Yogaraj Jayaprakasam, the company’s chief technology and digital officer, pictured.

So AI played a huge role in the IT modernization project, he says, using it to rewrite the old mainframe code, generate documentation, and regenerate code test cases. The company also used an AI-powered test automation suite, as well as AI tools to help move assets to the new cloud environment.

While AI can’t do everything during mainframe migration, it can make the move easier, Jayaprakasam says. “The biggest lessons learned is AI is one of the missing tools in your transformation tool set,” he adds.

Failing projects

Deluxe’s mainframe migration earned it a 2026 CIO 100 Award for IT innovation and leadership, but it also seems to have bucked a recent trend. Many mainframe migration projects haven’t gone as planned, and Gartner recently predicted that more than 70% of mainframe exit projects that started in 2026 will fail to produce intended benefits because of an overreliance on gen AI.

Mainframe transformation projects tend to work better when they’re part of a larger business transformation rather than a one-off project using gen AI to do most of the work, says Gartner analyst Alessandro Galimberti.

“Generative AI and agentic AI are extremely powerful, but they also have their own limits,” he says. “With all these kinds of tools trying to convert code or somehow fit into a non-mainframe workload, we don’t really see a track record of success.”

Gartner also sees a declining interest in mainframe migration projects, Galimberti says. With mainframes getting support from several IT vendors, and with a general lack of migration success, many companies are choosing to keep many workloads on their existing big iron.

But mainframes also have a proven track record of very high uptime and backward capability, Galimberti says.

“If I’m a bank, a financial institution, or a transportation company, I need to run applications that are the core of my business,” he adds. “I need reliability, transactional integrity, and security, and these applications have a low change rate over the years because they map very stable business processes.”

The Gartner prediction makes sense to John McKenny, senior VP and GM of Intelligent Z optimization and transformation for mainframe support vendor BMC Software.

“With organizations thinking about mainframe exits, the expected benefits they’re looking for are usually pretty straightforward,” he says. “They think, ‘It’s going to be lower cost, I’m going to get equal or better capabilities, I should be more agile.’  The reality is those outcomes rarely show up at scale.”

Mainframe migration is possible, but the successful projects tend to be small scale, McKenny adds. He was on a recent call about a failed migration project in Europe, with a large bank cancelling the project at the end of 2025 and recommitting to the mainframe as a strategic platform.

“I’ve never seen a large-scale mainframe migration project finish under budget, ever,” he says. “Most of the projects I hear about fail outright.”

Trying again

Like some organizations that Gartner has observed, Deluxe tried to move away from its mainframe several years ago, but the project failed, Jayaprakasam says. Yet the company took the steps it needed this time to ensure the new migration succeeded.

While a mainframe migration isn’t for every organization, the latest move made sense for Deluxe, he says.

The mainframe, after all, was the backbone for a large portion of the company’s annual revenue, and interfaces with several top banks across North America. The modernization effort rebuilt core business processes and data, moving them from the mainframe to a modern cloud-native technology stack, including Salesforce, Mulesoft, and SAP S4/HANA.

While AI played a big part, there’s danger in overestimating the power of AI during a migration project, Jayaprakasam says, and organizations need to follow best practices for IT migration.

“If you minimize the importance of communication, risk planning, and business alignment because you have AI, you tend to fail,” he says. “But as long as you play all those cards and recognize AI was the missing piece to the puzzle, then you have a much better chance of winning.”

Deluxe also used a cross-functional tiger team to look at the various options available to accelerate reverse engineering, including AI tools from OpenAI, Anthropic, as well as GitHub Copilot throughout the project.

In addition, AI was useful to dig through the mainframe code and understand what needed to be updated, Jayaprakasam says. Organizations sitting on decades-old code often no longer have people who understand it.

“I always tell people that the code remembers what the organization forgot, because with people going and changing, people don’t remember what we wrote in the code, but the code remembers,” he adds. “The amazing tool that was missing before is we didn’t have an interpreter who understood what the code remembered. Now with AI, you have the interpreter.”