Cut mainframe risk and cost by migrating one business area at a time
Many large companies still run critical systems on mainframes that frequently operate differently.
For example, in banking, account balance systems may remain unchanged for years, while services supporting new products and offers change every few weeks. Despite that divergence, leaders often choose between limited mainframe migration options—moving the whole 9 yards to new platforms or retaining the status quo.
Both approaches can be costly.
Migrating all systems at once risks disrupting reliable operations for little gain. Keeping the present setup forces rapidly changing services to follow inappropriate update cycles designed for stable systems.
Gartner predicts that more than 70% of projects started in 2026 to move companies off mainframes will fail to realize their intended benefits.
A more efficient mainframe migration strategy is to migrate one business area (domain) at a time, giving priority to those that need frequent updates and leaving settled domains unchanged. Progress should be measured by the speed of business change and systems operating costs.
Why the exit-or-stay debate stalls
Gartner attributes much of the anticipated failure to unrealistic expectations of generative AI. Analysts note that current tools are better at enhancing existing code than at migrating decades-old business logic to new platforms. For medium-sized environments (the largest market sector), Gartner recommends limiting full exits to specific cases and aligning each workload with the most suitable platform.
Standing still has its own price. In 2025, The U.S. Government Accountability Office reported that federal agencies typically spend about 80% of their IT budgets on operating and maintaining existing systems. Additionally, eight of the eleven most critical legacy systems reviewed still depend on outdated programming languages.
If neither approach works, the key question becomes how to set appropriate boundaries.
The exit market is heading for a reset
Gartner's second prediction argues against staking everything on one long program. By 2030, it expects 75% of vendors in the mainframe exit market to change their business models or stop operating. A multi-year exit signed in 2026 could outlast the firm contracted to deliver it.
That makes partner durability a board-level concern and favors plans where each phase pays for itself before the next begins.
Measure success by change speed and run cost
Many modernization programs report progress by retired capacity. While this is easy to measure, it doesn't indicate how quickly the business can deliver changes or the cost per transaction.
This gap is important because migrating code to a new platform without redesigning it rarely improves delivery speed. Rehosting can provide cost savings plus operational consistency, but these benefits should be considered as cost reductions.
A more efficient metric is change lead time, which DORA measures from code commit to production. Paired with run cost per transaction, it shows whether modernization is creating value across platforms. DORA's 2025 research cautions that AI amplifies existing strengths and weaknesses, so speeding up the conversion of a poorly understood environment is likely to accelerate its issues too.
At American Airlines, the mainframe-based Flight Operating System (which manages crew and flight schedules) stayed on its original platform. DXC and the airline integrated it into American's DevOps toolchain, so progress is now measured by delivery performance rather than platform location.
Prioritize domains by frequency of change
Measured this way, the mainframe migration approach typically divides estates into three categories:
- Keep and expose. Stable, high-volume transaction modules typically perform best in their current environment, exposed via APIs to manage costs effectively. For instance, banks using DXC's Hogan core banking platform can now offer AI-personalized installment plans through a partnership with Splitit, without replacing their core systems. Similarly, Northeast Grocery extended its mainframe’s life instead of purchasing a new one, saving over $40 million in 5 years.
- Move one at a time. Domains facing ongoing digital demands, where speed is a key differentiator, are prime candidates for migration. A leading North American insurer moved annuity administration from the mainframe to DXC's insurance platform on AWS, migrating more than 1 million contracts in 8 months. This reduced operating costs by 40% and improved time to market for new services by 15%.
- Relocate for cost and resilience. Some domains have reliable code but are hosted on suboptimal platforms. The IT services arm of a Munich-based insurer moved its mainframe systems to a DXC data center as part of a hybrid model, completing the transition within 8 months without service disruption.
Five questions typically determine which mainframe migration options suit each domain best: “How often does it change?”, “Does it differentiate the business?”, “How tightly is it coupled to other systems?”, “What do the regulators require?” and “Do the support skills still exist?”
Prioritize using AI for understanding
Gartner's caution doesn't exclude AI. Its most reliable value today is in comprehension, such as documenting unwritten business rules, mapping dependencies and generating tests. These activities deduce risk regardless of whether a domain is retained or replaced.
DXC's Modernization as a Service (MaaS) uses AI to analyze and convert legacy code, with engineers verifying that the converted code functions identically to the original. Through its June 2026 alliance with Anthropic, DXC uses Claude to analyze, refactor and transform legacy codebases.
Plan to run old and new side by side
This hybrid mainframe migration approach often spans several years, making continuity across both environments essential. DXC manages more than 140 IBM Z platforms and has completed more than 250 mainframe migrations. ISG's 2026 Provider Lens reports recognize DXC as a leader in mainframe services across the U.S., Europe and the U.S. public sector. DXC OASIS supports this transition by combining human expertise with agentic AI to quickly identify operational risks.
Draw the domain map before setting an exit date
Mainframe migration strategy works best when made at the point of business value creation. Over the next quarter, decision makers should map domains by change frequency and run cost, establish baseline change lead time and cost per transaction, migrate one fast-changing domain and expose one stable core through APIs. Leaders must then ask each potential partner if it can manage both sides throughout the transition.
This approach delivers value in stages, safeguards critical operations and maintains flexibility as the vendor market evolves.
