Most banks have invested in AI by now - digital channels are live, customer-facing journeys have been redesigned, and automation has been applied wherever it was visible and measurable.
And yet, operational costs keep climbing, headcount grows with volume, and every new initiative seems to cost more to deliver than the last.
Leadership keeps returning to the same question: where is the money actually going?
Most of it is going somewhere that doesn't appear on any dashboard.
The investment has been real, and so has the problem.
Banks now spend approximately $600 billion a year on technology globally, growing at 9% annually - well above revenue growth of 4%. Despite this, labor productivity at US banks has been falling 0.3% a year since 2010, even as most other sectors have posted gains.
The correlation between a bank's revenue and its headcount remains stubbornly high. The industry has not been able to buy its way out of a linear relationship between growth and people.
Even though banks have been measuring system performance, the cost that keeps rising lives outside every system they measure.
The work in banks nobody is measuring
Every operational dashboard is built around system performance - transactions processed, cases opened, applications submitted, and calls handled. What those dashboards don't capture is the work that happens between those events.
Take a disputed transaction. The customer flags it through the mobile app. Someone in operations pulls the transaction record from the payments system, checks the fraud system for any associated alerts, opens the customer record in a separate platform to verify history, requests supporting documentation through a third tool, and then manually summarizes the findings into a case that gets routed to a resolution team. Each of those steps is performed by a person, in a different system, with no shared view of what the others have already done.
That coordination work doesn't appear in any system's performance report because no system owns it. It lives in the whitespace between systems - in inboxes, in spreadsheets, and in calls between teams working from different views of the same customer.
This isn't an edge case. Estimates consistently put this kind of inter-system coordination work at around 50% of all frontline activity. It is the most expensive category of work in the operation, charged at fully-loaded employee cost, and it is invisible to every efficiency metric banks track.
This is the cost that doesn't show up - until someone tries to find out why headcount keeps growing despite years of automation investment.
Three cost categories that never appear in operational reporting
The whitespace generates cost in three distinct ways, none of which typically show up in standard operational metrics.
Coordination cost is the most direct. It's the staff time spent moving work between systems - pulling data from one platform to populate another, chasing status across teams, and reconciling records that should already be consistent. This work is high volume, performed by trained employees at full salary cost, and it produces no customer value.
Delay cost is less visible but significant. Every handoff between disconnected systems adds time to a resolution. A dispute that could close in one interaction takes multiple touchpoints because the resolution team doesn't have the full picture the customer service team already built. A loan modification that should move in hours takes days because the credit system can't see what the document system already received. That delay has a cost in customer satisfaction, in abandonment rates, and in the operational overhead of managing open cases across longer resolution windows.
Initiative cost is the one that surprises most leadership teams. When every new capability requires custom integration work to connect it to adjacent systems, the unit cost of change stays permanently high. Banks find that new initiatives are expensive to deliver because the same integration effort gets rebuilt from scratch every time. They pay more because the foundation requires the same effort to build on every single time something new is added to it.
Why adding more technology hasn't fixed this
The instinct when costs persist is to add more capability - better tools, more automation, and AI layered onto existing processes. Each new tool does what it's designed to do, but lands on the same disconnected foundation and inherits the same coordination problem.
The whitespace doesn't shrink - it reorganizes. Every new capability requires its own data feeds from adjacent systems, its own mapping to upstream and downstream processes, and its own workaround for the gaps between them.
AI can automate work that lives inside a system, but it cannot reach work that lives between systems - and that's precisely where most of the cost is. For example:
- Automating dispute resolution requires consistent customer data from across the operation.
- Automating loan exception handling requires structured workflows that behave the same way regardless of which channel originated the application.
- Automating KYC remediation requires a shared view of what has already been verified and where. None of those conditions exist on a fragmented foundation by default.
Every tool deployed without fixing the foundation simply adds to the number of systems the coordination work has to bridge. This is why adding technology hasn't bent the cost curve.
What changes when the gaps close
Banks that have connected their operation into a unified frontline - digital channels, front office, and back-office working from a shared foundation - don't just report lower cost-to-serve. They also eliminate entire categories of work.
The coordination cost disappears because there is no coordination to do. A customer interaction that begins in a mobile channel carries its full context into the branch and into the operational case that follows without manual transfer, reconciliation, or calls between teams.
The delay cost shrinks because resolution teams have the complete picture from the moment a case opens. The initiative cost drops too, because new capabilities connect to a foundation that already provides the data, workflows, and governance they need.
The whitespace doesn't get cheaper to manage. It gets eliminated. And the banks on that side of the divide are compounding the advantage every quarter. Meanwhile, those still paying the coordination tax keep absorbing the same cost on every interaction, every exception, and every new initiative they try to deliver.
.jpg)
