The hidden cost of legacy systems: up to 40% of your IT budget is already propping up the old

Jorge García

Tecnea

The hidden cost of legacy systems: up to 40% of your IT budget is already propping up the old

Between 21% and 40% of what your company spends on technology builds nothing new: it goes toward propping up systems that, in many cases, should have been replaced years ago. This isn't an alarmist estimate from a vendor with something to sell — it's what more than 660 technology leaders (CIOs, CTOs, CISOs, CDAOs) worldwide report in Deloitte's latest major technology leadership study.

Management team reviewing technology spend charts on a laptop For every $100 spent on IT, between $21 and $40 goes toward sustaining what already exists, according to Deloitte.

The figure that changes the question

Deloitte's Global Technology Leadership Study 2026 surveyed more than 660 technology leaders at organizations worldwide between December 2025 and February 2026. Its conclusion on technical debt is blunt: for every $100 an organization invests in IT, between $21 and $40 goes toward sustaining what already exists — patches, workarounds, brittle integrations — instead of building something that creates new value.

The question that actually matters isn't "how much does modernizing cost?" It's "how much is not modernizing already costing me?" — and according to the study, that second figure is usually larger, and grows every year it's deferred.

Eight signs your software has already become the problem

A report by consultancy Excelia, covered by MuyPymes, lists eight symptoms of an obsolete ERP (or any central management system). They're easy to recognize because, if you've spent any time inside your company, you've probably lived through them firsthand:

Difficult or impossible updates — the vendor no longer offers support, or every patch is a project in itself.
Poor integration — it doesn't connect with your CRM, website, or other tools without fragile custom workarounds.
Manual processes — what should be automated is still handled with spreadsheets and duplicate data entry.
Limited mobility — no cloud version, no useful remote access, no real use from a phone.
Unreliable data — reports that arrive late, incomplete, or wrong.
High maintenance costs — infrastructure and specialist support eating up budget year after year.
Poor user experience — outdated interfaces that create internal resistance to using the tool at all.
Lack of scalability — it can't keep up with business growth or regulatory change.

Objectivity note: Excelia is a consultancy that sells, among other services, ERP modernization — it has a direct interest in these symptoms being seen as urgent. The list itself, however, is recognizable regardless of who publishes it. The same article cites a Microsoft data point — another party with a direct commercial interest, as the seller of the cloud platforms that replace these systems — putting the cost of maintaining an obsolete ERP at up to 90% of a company's IT budget.

Why AI changes the calculus of touching it — with caveats

For years, the main reason not to modernize a legacy system was cost and risk: understanding what code written 20 or 30 years ago actually does, undocumented and with the original developer long retired, could take months of analysis before a single new line got written.

On February 23, 2026, Anthropic announced an AI capability that can automate much of that upfront work: reading legacy code (the most-cited example is COBOL, still running inside banking, insurance, and government), mapping its dependencies, tracing execution flows, and documenting what each module does — the phase that, according to Anthropic, accounts for most of the cost and time of a modernization. According to the company, work that used to take years can now shrink to quarters.

Close-up of a circuit board with old electronic components Understanding what a legacy system actually does, with no documentation and no original team, has always been the most expensive part of modernizing it.

The market took it seriously: that same day, IBM's stock fell 13% — its worst trading day since 2000, wiping out more than $31 billion in market value — dragging Accenture and Cognizant down with it, as CNBC reported. The reason: a large share of these companies' system-modernization business is built precisely on that upfront, consulting-hour-intensive work AI now promises to shrink.

Objectivity note: Anthropic sells AI tools for software development, so it has a direct commercial interest in this narrative. Worth adding IBM's response, reported by IT Pro: "decades of hardware-software integration cannot be replicated by moving code." This isn't magic or a single button — expert judgment is still needed to decide what to modernize, how, and with what guarantees. What's changed is that the analysis phase, previously cost-prohibitive, no longer is.

What this means when you don't have an in-house IT team

Almost everything above was written with large banks and insurers running thousands of lines of COBOL in mind. But the underlying mechanism — old systems that can no longer keep up with what the business needs, now understandable and modernizable much faster with AI — applies just as well to a mid-sized company with no IT department: the ERP bought ten years ago, the spreadsheet acting as a database because "that's how it's always been done," the custom software from a vendor that no longer exists.

A February 2026 Gartner report on manufacturing CIO challenges — summarized by Cora Systems, a project management software vendor with a direct interest in selling the alternative — identifies accumulated technical debt as one of three main headwinds for 2026: aging IT, OT, and engineering technology systems weren't designed for machine learning, predictive analytics, or automated decision-making — so AI pilot projects stall before they can scale, not for lack of ambition, but because the foundation they're built on can't hold.

It's the same conclusion, applied to a different sector: it's not just about the cost of maintaining the old. The old also blocks the new.

What to do about it

There's no need to rush into replacing every system at once — that's exactly the kind of hasty decision that usually ends up expensive. A reasonable order:

Audit before deciding. Check the eight signs listed above against your critical systems: how many apply, and since when.
Calculate the cost of doing nothing, not just the cost of modernizing. Hours lost to manual processes, opportunities not pursued because the system won't allow it, accumulated security risk.
Tell patching apart from modernizing. A patch buys time; it doesn't fix the root cause. It works as a bridge, not as a strategy.
When you decide to modernize, evaluate custom AI-based development as a real option — not just the traditional years-long migration with an army of consultants. The analysis phase, previously the most expensive part, is exactly what's gotten cheaper.
Don't leave the decision to whoever's selling the solution. Most of the data in this article comes from parties with a commercial stake in one outcome or another — start by understanding your own situation before listening to anyone's sales pitch, including ours.

Frequently asked questions

What exactly is "technical debt"? It's the accumulated cost of maintaining systems built on old decisions — technology, architecture, or shortcuts — that no longer fit what the company needs today. Like financial debt, it accrues "interest": the longer it goes unaddressed, the more expensive it becomes to fix.

Is this only a problem for very large companies with mainframes? No. The mechanism is the same for a mid-sized company running a ten-year-old ERP or a process that depends on a spreadsheet: the longer it goes untouched, the more expensive and risky it becomes to fix, and the more it limits what the company can do with AI or any other new technology.

Can AI modernize my legacy system on its own, with no human involved? No. What's changed is the analysis and documentation phase — understanding what the current system does — which used to consume most of the time and cost. Deciding what to modernize, how, and with what guarantees still requires expert judgment.

Where do I start if I don't have an IT department? With an honest audit: which systems your company actually runs today, how many of the eight signs above apply, and what real cost — not just financial — comes from leaving them untouched.

At Tecnea we build custom AI-powered applications that replace obsolete processes and systems with something designed for your company today, not for the one it was fifteen years ago. If you want to know what your current situation is really costing you, calculate your savings with AI.

This article is informational and does not constitute technical or investment advice. We have tried to indicate the source of each figure and, where applicable, whether whoever published it has a commercial interest in the outcome.