Why Banks Still Struggle to Turn Billions in AI Spending into Value
By Matthew Maunder, Founder & Principal, Valent Advisory
Simple Subscribe
Subscribe Now!
Banking has entered a harder phase of the AI cycle. Pilots are reaching production and employees are finding ways to use the technology. Now comes the uncomfortable question: where is the value?
KPMG’s Q1 2026 U.S. banking survey puts average projected AI investment over the next 12 months at USD 177 million, while Cambridge found that 76% of large financial institutions struggle to measure AI value.
Key insight: The biggest risk is no longer that AI fails. It is that AI works — but the value never reaches the business.
Need to Know:
- AI activity is not AI value. Pilots, users and time saved do not prove that a business outcome has changed.
- Workflow change is necessary but not sufficient. A better process can still create capacity that never reaches the business.
- Freed capacity needs a destination. Leaders must decide whether it supports growth, absorbs volume, avoids hiring or reduces cost.
- “Every material initiative needs a value realization gate”. The next investment decision should depend on evidence, not momentum.
AI Has Moved from Experimentation to Economic Accountability
Banks are putting hard numbers against AI. RBC is targeting CAD 700 million to CAD 1 billion in incremental enterprise value from AI by 2027, net of investment. DBS reported approximately SGD 1 billion in economic value from its data analytics and AI/ML initiatives in 2025.
JPMorgan Chase is increasingly framing AI around tangible outcomes. In its 2026 Company Update, the bank said it was seeing benefits across revenue and expense and had doubled the number of use cases in production. On its Q1 2026 earnings call, CEO Jamie Dimon cautioned that AI productivity does not automatically translate into a structurally better efficiency ratio as competitors adopt the same technology.
These are no longer innovation ambitions. They are business-value commitments.
Why this matters: The conversation is shifting from What can we deploy? to What value did we actually realize?
- Separate experimentation metrics from value metrics.
- Track which use cases scale and what changes in the business.
- Make the economic or customer outcome visible from the start.
Start With the Business Problem and Redesign the Work
Technology investments lose value when treated as technology programs first and business transformations second.
Start with the operating problem. Business leaders know where work is slow, fragmented or manual. Technology should remove those roadblocks to achieve a defined outcome. The strongest model is business-led and technology-driven.
Deployment still is not enough. Employees can adopt a platform while preserving the spreadsheet workaround, legacy approval or manual handoff it was supposed to eliminate. At that point, the organization has built a very expensive filing cabinet.
Consider an AI assistant that cuts customer-issue research from 12 minutes to 8 minutes. If routing, escalation rules and staffing assumptions remain unchanged, the bank may record productivity without capturing much economic value.
The technology worked. The operating model did not.
Why this matters: Workflow redesign is a bridge to value, not the finish line.
- Define the business problem and baseline before selecting or scaling the solution.
- Design the future-state workflow as part of the investment case.
- Remove the workaround or duplicated activity the technology is meant to replace.
Productivity Is Not Value Until Someone Captures It
If AI turns a 60-minute task into a 20-minute task, what happens to the other 40 minutes?
That is where theoretical productivity can become stranded capacity.
Citi reported that AI-assisted coding tools are creating approximately 100,000 hours of capacity each week, allowing developers to focus on higher-value innovation. The broader lesson: capacity needs a destination.
My bias is toward redeployment and growth where the economics support it — more client time, greater volume or higher-value work. In a mature business, taking cost out may be more disciplined. Cost avoidance can create real value, but it is not the same as direct expense reduction.
Why this matters: Time saved is an operating benefit. The management decision that follows determines whether it becomes enterprise value.
- Decide whether released capacity will support growth, absorb volume, avoid hiring or reduce cost.
- Track hard savings, cost avoidance, and redeployed capacity separately.
- Do not count a financial benefit until there is evidence the capacity was captured.
Make the Business Own the Benefit
Many programs have an executive sponsor. That is not the same thing as having a benefit owner.
If an AI initiative promises lower cost-to-serve, increased sales, faster throughput or a better customer experience, the business executive accountable for that outcome should also own its realization.
Technology should own enablement, architecture and controls. The business should own the outcome and value realization. Finance can help distinguish forecast benefits, avoided costs and realized impact.
Why this matters: Delivery accountability gets the capability live. Benefit accountability turns it into a business result.
- Assign a named business owner to each material benefit.
- Define how the benefit will be evidenced and validated.
- Review realized value through the same cadence used to review business performance.
Put a Value Realization Gate Before the Next Dollar
Banks need a deliberate point when experimentation becomes an investment decision.
A “value realization gate” asks whether the original business hypothesis is being proven before more capital and talent are committed. Timing should reflect the use case and complexity, but the gate should be established before the pilot begins.
Business problem → workflow change → adoption → capacity decision → operating outcome → financial or customer value
If leaders cannot trace the benefit through that chain, they should be cautious about calling it realized value.
Why this matters: Value realization should be designed into the initiative, not calculated after the fact.
Before approving the next investment, leaders should answer four questions:
- What outcome are we buying? What revenue, cost, risk or customer result should change?
- What changes operationally? Which work disappears or improves, and what happens to the capacity created?
- Who owns realization? Which business executive is accountable for converting the improvement into a measurable result?
- What evidence earns the next dollar? What would justify scaling, extending, redesigning or decommissioning the initiative?
Bottom line: Banks have demonstrated that they are willing to spend on AI. Increasingly, they are putting billion-dollar expectations against the return.
The challenge now sits in the space between the two.
Technology has to change the workflow. Workflow improvement has to create an operating benefit. Capacity has to be deliberately captured. That benefit then has to reach a financial or customer outcome the business can defend.
The winners in banking’s next phase of AI will not be defined by how much technology they deploy.
They will be defined by how much value they actually realize.
