Most banks focus on AI potential but overlook the hidden operational, architectural, and governance costs that erode value long before AI generates measurable returns.
Your AI strategy has a tax. Most banks don’t know they’re paying it.
There is a number that should be on every bank board agenda right now. It is not a revenue figure. It is not a capital ratio. It is not an NPS score.
It is 94%.
94% of core banking modernization projects exceed their original timelines. — IBM Institute for Business Value, 500+ banking CIOs surveyed, Sept. 2025
That is the share of core banking modernization projects that exceed their original timelines, according to IBM’s Institute for Business Value, based on a global survey of more than 500 banking CIOs and nearly 200 Chief Data and AI Officers at institutions with over $10 billion in assets. Not community banks struggling with constrained budgets. Not outliers. The mainstream.
Ninety-four percent.
Accenture’s 2026 Banking Top Trends report found that 61% of bank executives plan to invest directly in AI this year. The boardroom mandate is real. The infrastructure question is whether the core beneath it can support what is being built on top.
The Tax No One Is Naming
When a core modernization project runs over, the cost does not show up cleanly on a P&L. It spreads. It hides in extended vendor contracts. In consultant scope creep. In the parallel IT teams running the old system while the new one catches up. In the AI initiatives that get queued, deprioritized, and quietly deferred because the data infrastructure they depend on is not ready. That is the tax. Not a line item. A drag. Structural, persistent, and almost never attributed correctly to its source.
Accenture estimates that nearly 70% of bank IT spending today goes toward maintaining legacy systems and meeting regulatory requirements. Not building. Maintaining. That leaves roughly 30 cents of every technology dollar for everything else, including the AI your board approved last quarter.
“A regional bank found itself devoting most of its annual IT budget to patching legacy systems. No funds remained for modernization, until the CFO aligned the budgeting process with the bank’s stated goal of shifting 25% of IT spend to digital transformation within two years. “ — Rick Mavrovich, The Strategic Flywheel: How to Run, Change and Innovate the Bank, 2nd Edition
That bank is not an outlier. It is the pattern. The CFO in that story did something most do not: named the problem on the balance sheet instead of in a hallway. That is the first move. Without it, the tax stays invisible.
What I Have Seen From The Inside
I have sat across the table from bank executives who are genuinely convinced their AI strategy is on track. They have the roadmaps. They have the vendors. They have the board approval.
What they do not have, and often do not yet know they do not have, is a core that can actually run what they are building on top of it.
The legacy core is not a technology problem. It is a decision architecture problem. The decisions that created it were rational at the time. Batch processing made sense when transactions settled overnight. Siloed data made sense when regulatory reporting was the primary consumer. The problem is not that those decisions were wrong then. The problem is that they are catastrophically expensive now, and the cost compounds every quarter that AI sits in the queue waiting for clean data it cannot get.
Less than half of banking CIOs in the IBM IBV study reported achieving meaningful gains from their modernization investments. Some reported that operational capabilities actually got worse during the transition. This is not a vendor problem or a project management problem. It is a structural reality that the industry has been reluctant to state plainly.
The core IT infrastructure of a bank, as I wrote in The Strategic Flywheel, is “the invisible engine in the operation that allows it to run and move on. ” Invisible is exactly right. And invisible problems do not get fixed. They get budgeted around, year after year, until the cost of avoidance exceeds the cost of the cure.
Why 2026 Is Different
Previous years, the pressure was theoretical. Boards asked about AI readiness in the abstract. Regulators mentioned legacy technology risk in passing. Vendors pitched modernization without urgency. That has changed.
The OCC named legacy core systems as a 2026 examination focus. IBM IBV found that only 9% of banks with over $10 billion in assets are live or ready to deploy key AI and tokenization initiatives this year. The gap between where boards think their institutions are and where the infrastructure actually is has never been wider, or more consequential.
The banks that figure this out in 2026 will have a compounding advantage. Every modernization decision made now shortens the queue for the AI initiatives behind it. The banks that do not will still be paying the tax in 2028, just at a higher rate.
Rick’s Take
The 94% number does not surprise me. What surprises me is that it is not driving more urgency at the board level.
The reason it does not is that the tax is invisible in the normal reporting cadence. Project overruns get reforecast. Delayed AI initiatives get rescheduled. The core sits underneath everything, consuming budget, constraining optionality, and never appearing on a single dashboard as the root cause.
My advice to any bank executive reading this: pull your last three major technology initiatives. Look at what caused the delays.
I will tell you what you are going to find.
