Legacy Modernization Economics: The Business Case for AI-Accelerated Modernization

As much as 70 percent of the software running Fortune 500 companies was built 20 or more years ago, according to McKinsey research, and technology now enables roughly 71 percent of the value generated by business transformations. Those two facts sit next to each other uncomfortably: the systems most enterprises depend on for growth are also the oldest, most expensive systems they own.

Legacy modernization has been treated as a discretionary IT problem for decades, one business leaders were content to defer. AI does not remove the case for caution. It changes the actual cost and speed of modernizing, which means the math that justified deferring the decision no longer holds the way it used to.

The reasons legacy modernization keeps getting pushed down the road are familiar to any CIO. It is too expensive, often running into the hundreds of millions of dollars. It takes too long, typically five to seven years for a major system. It is too disruptive to a business that depends on the system working every day. The return on investment is hard to pin down in advance, and the current system, for all its flaws, basically works.

What that framing leaves out is the cost of not acting. McKinsey’s research found that at a large European bank, 70 percent of IT capacity was spent simply maintaining legacy systems, capacity that never went toward anything the business could point to as new value. More broadly, technology debt can account for up to 40 to 50 percent of total technology investment spend once its effect on the P&L is properly measured, and companies pay an additional 10 to 20 percent on top of the cost of any new project just to work around the debt already on the books.

The scale of what runs on these systems raises the stakes further. Legacy platforms in financial services and insurance handle transactions worth trillions of dollars daily and administer policies representing trillions more in annual premiums. Deferring modernization on infrastructure at that scale is not a neutral choice. It is a compounding one.

The change AI brings is not faster typing. It recalibrates the cost and timeline that made modernization prohibitive in the first place.

McKinsey’s research offers a concrete before-and-after: a transaction processing system for a leading financial institution that would have cost well over $100 million to modernize three years ago now costs less than half that using generative AI. Across early programs, McKinsey found gen AI-driven modernization delivering a 40 to 50 percent acceleration in modernization timelines and a 40 percent reduction in costs tied to technology debt.

The mechanics show up clearly in specific engagements. At one bank attempting to modernize a mainframe, a project estimated at 700 to 800 hours to migrate 20,000 lines of code saw that estimate cut by 40 percent once an orchestrated set of gen AI agents took on the work, with the relationship-mapping step alone dropping from 30 to 40 hours down to about five. At a top 15 global insurer, reverse-engineering legacy code with gen AI agents to understand it before rebuilding it delivered more than a 50 percent improvement in modernization efficiency and testing, and more than a 50 percent acceleration of coding tasks.

Ascendion’s own delivery record shows a similar pattern at a different institution. For a digital-first banking pioneer, AAVA™ reverse-engineered more than 900,000 lines of 1980s code in three weeks, part of a modernization program that landed at roughly a quarter of the original $36 million projection and in half the time a traditional approach would have taken.

None of this value is automatic, and McKinsey’s research flags a specific failure mode worth taking seriously: using gen AI to translate old code directly into a modern language, line for line, without changing what the system actually does. That approach, sometimes called “code and load,” does not eliminate technical debt. It relocates it into a newer, more modern-looking context, the same trap many enterprises fell into during the early days of cloud migration, when “lift and shift” moved legacy problems onto new infrastructure without solving them.

A modernization business case built on lines of code converted, rather than business outcomes achieved, will show activity without showing value. The alternative McKinsey’s research points to is using gen AI first to understand what a legacy system is actually doing, in plain language, then deciding with business stakeholders what needs to be modernized, what needs to be redesigned, and what can simply be discarded. That distinction, between translating debt and actually resolving it, is what separates a modernization program that pays for itself from one that just moves the bill.

Run this correctly and the return shows up in specific, financeable line items rather than vague productivity claims. Reverse-engineering a legacy system properly, rather than converting it blindly, surfaces hidden technical debt that was never visible on a balance sheet. For one financial client, that discovery process uncovered $12 million in hidden technical debt in just four weeks, giving the business a number to prioritize against instead of a guess. Rebuilding on modern, AI-orchestrated workflows also compresses time to market directly: one media company using this approach launched a new streaming platform six months ahead of schedule. And streamlining the workflows around the legacy system itself, not just the code, can cut cycle times by as much as 60 percent, which is where the ongoing savings compound well past the initial modernization project.

A credible modernization business case rests on a few specific moves, not a general appeal to AI’s potential.

Start by sizing the run-cost baseline, what share of current IT capacity or budget is already going toward legacy maintenance. That number, not the cost of the modernization project itself, is usually what flips the decision from “too expensive to modernize” to “too expensive not to.”

Anchor the case on business outcomes rather than code volume, to avoid the code-and-load trap. A plan that measures success in lines of code converted will optimize for the wrong thing.

Scope the effort at the domain level, an end-to-end business process, rather than system by system. McKinsey’s research found that as many as 80 percent of successful interventions in a struggling digital transformation are based on re-anchoring the scope around a well-defined domain rather than an isolated technical fix.

Track value continuously rather than only at the end of the project. A modernization program that only reports progress at close tends to quietly default to delivering code instead of the value it was funded to deliver.