Banks don’t have an AI funding problem. They have a cost allocation problem.
While most banks are investing heavily in AI, many are struggling to move beyond pilots and isolated use cases. The challenge is rarely a lack of opportunity. It is an inability to release and reallocate capital fast enough to support meaningful transformation.
The institutions generating the greatest returns from AI are not simply investing more. They are removing structural cost from the organisation and redirecting that funding into AI-led change.
This requires two steps.
1. Taking a zero-based view on the current cost base
Funding AI starts with releasing cash from within the business. Not through marginal efficiencies, but by going after areas where spend is no longer aligned to value. In most banks, parts of the cost base exist simply because they always have. Low-value activities, such as legacy reporting, continue to absorb funding without a clear link to outcomes, quietly inflating the cost base. They are widely recognised internally but rarely removed.
Third party spend is one of the clearest examples. Multi-year contracts for core banking platforms, market data feeds, cloud consumption and outsourced KYC/AML utilities are rarely revisited once signed, and business lines often contract independently with the same vendors, creating duplication that a more coordinated approach would remove. A long tail of small, low-value supplier contracts adds further drag, consuming disproportionate procurement effort relative to the value it delivers. In many banks, procurement still operates as a transactional, purchase-order function rather than active category management, so this spend goes largely unchallenged.
The starting point is simple: identify where work exists without a clear outcome and stop doing it. Not optimise it, not automate it – stop it. If it does need to be done, ask whether it can be delivered differently. In practice, this means eliminating low-value activities, collapsing duplicated work, consolidating fragmented vendor contracts, and only then looking at lower-cost delivery models or outsourcing.
In practice, these conversations often don’t happen. There’s no real forum to challenge spend, value is poorly quantified, and perceived risk is exaggerated, often to protect the status quo.
What changes the dynamic is objectivity. Looking at the cost base through the lens of an external investor, without legacy constraints or internal bias, quickly surfaces where spend isn’t justified. That creates the momentum and confidence at executive level to act on more structural opportunities. From there, it becomes operational. Approaches like zero-based budgeting reset the baseline, asking what it would cost to run the business at a legal and regulatory minimum, and rebuilding from there. It forces a much clearer view of what each pound is actually delivering.
The fix: Force a zero-based challenge of the cost base – including third-party and vendor spend –remove activities that no longer justify their existence and ringfence that funding for AI.
2. Redesign the operating model for AI, and resize around it
Funding AI is not just about cost reduction. It requires changing how the organisation is structured to deliver work. AI doesn’t simply automate tasks; it removes steps, reduces hand-offs, and concentrates work into fewer, higher-value activities. That requires a shift in the operating model, not incremental adjustments to existing teams.
In most cases, AI is layered onto existing structures. The result is predictable: fragmented use cases, duplicated effort, and benefits that never fully materialise. The alternative is to design an AI-enabled operating model, one that reflects where work is genuinely required, where it can be automated, and where capability should sit.
This starts with how the bank faces the market. Many institutions are still organised around products, regions, or legacy business lines, each with their own teams, processes, and duplication of effort. Procurement is a good example: category managers and supplier relationships are frequently replicated across divisions rather than centralised, even where the underlying spend and suppliers overlap. AI creates an opportunity to rethink that, centralising capabilities where it makes sense, sharing scarce expertise, and removing duplication. It also changes how support functions operate. In procurement, operations, and finance, AI can absorb large portions of transactional and analytical work, allowing entire layers of activity to be consolidated or removed.
Procurement also shows what this looks like in practice. AI can take on much of the contract analysis and redlining, supplier risk screening, spend classification and tail spend management, and sourcing support that currently consumes significant analyst time, freeing the function to focus on strategic sourcing and supplier relationships. The same applies to third party risk management, where AI can absorb much of the manual due diligence and ongoing monitoring that has grown heavier under recent operational resilience and outsourcing requirements. Because procurement is rarely customer facing or regulated in the way risk and compliance are, it is often one of the easier functions to centralise and resize first – and the savings can be ringfenced directly to fund AI investment elsewhere.
But this only works if the organisation follows through. If the operating model doesn’t change, the same work continues in the same way, just with more technology layered on top.
The fix: Design the operating model around how AI changes work, how it flows, where capability sits, and how resources are shared. Then resize the organisation to match. If headcount and structure don’t change, the cost base doesn’t change.
AI doesn’t require new funding. It requires removing cost that no longer earns its place.
Banks that scale AI successfully are the ones willing to be bold and make trade-offs, taking cost out and reshaping their organisation to fund what comes next. Those that don’t see the same outcome: pockets of activity, limited impact, and a growing perception that AI is expensive without delivering real value.
The constraint is not the technology. It is the willingness to make structural change.
This is already being delivered in practice, combining cost transformation with AI adoption to create a self-funding model. At 4C Associates, this means identifying where cost can be removed, linking it directly to AI investment, and ensuring the organisation follows through. Without that, AI remains a cost. With it, it becomes a structural advantage.
The banks that win with AI won’t necessarily be those that invest the most. They’ll be the ones that remove cost fastest and redirect it towards change.
If you’d like to discuss how to unlock funding for AI through structural cost reduction and operating model redesign, we’d be happy to share our experience: Andy Hemsley, Rohit Namdeo, Agnieszka Abbott MCIPS, Sebastian Kay.
