As AI is moving beyond the pilot stage, the question arises whether banks can turn growing AI consumption into results that are predictable, repeatable and defensible.
The numbers suggest most cannot yet. The 2026 State of Tokenomics study of 472 organisations found that three in four could not confidently prove AI business outcomes to the CFO.
Closer to home, 54% of Asia-Pacific respondents in an IDC study for Expereo reported cost overruns or missed ROI, 80% in Malaysia.
In a Cognizant survey of 250 FinOps leaders, 72% had exceeded their AI budgets in the past year.
The issue isn’t the price of a token. It’s the operating system around AI: who consumes it, for which outcome, under what policy, on which model, and with what evidence of quality and return.
Visibility is the first control
AI spend enters a bank through many routes, from AI model APIs to embedded software features and self-hosted infrastructure. The model invoice is only part of the bill: vector databases, embeddings, orchestration, observability and data movement can account for 40% to 60% of an AI feature’s cost.
That makes a monthly token total a weak management metric.
Cost per onboarding file processed, servicing request resolved or alert reviewed tells a bank far more than cost per million tokens. Failed and discarded runs must stay in the denominator.
Buyers already know this: 23% of State of Tokenomics respondents asked providers for more granular usage data. Only 4% asked for lower prices.
Control must follow every request
Once consumption is visible, it can be steered. Not every task needs a frontier model—and one can cost 50 to 100 times more per token than the smallest model from the same provider.
Routing, the practice of dynamically directing an incoming prompt or task to the most suitable AI model, is going mainstream: 86% of respondents use or are evaluating a model router, and users are four times more likely to show the CFO measurable value.
But routing isn’t just a cost lever. It’s a resilience and governance control. Sensitive workloads stay inside the enterprise. Traffic moves when a provider slows or fails. Policy travels with every request, rather than relying on users to choose correctly.
Governance must produce evidence
For banks, these disciplines are converging with regulation. The Monetary Authority of Singapore’s AI risk management guidelines, published on 7 October 2026, call for clear accountability, an accurate inventory of AI use, risk-materiality assessments, lifecycle controls and oversight of third-party AI.
Governance expectations apply from October 2027, and lifecycle controls from October 2028.
A bank cannot keep a reliable inventory when AI activity is fragmented across providers. It cannot assess materiality without knowing the model, data, owner and purpose behind each use case. And it cannot evidence oversight without a record of approvals, changes, performance and exceptions.
Agentic AI raises the stakes: 97% of organisations in Cognizant’s FinOps survey had an AI agent session escalate beyond expected cost or activity. MAS plans to consult on agentic AI in 2027.
Governance can’t be a gate at the end of development. It has to run continuously, in the flow of work.
The same evidence that satisfies auditors and regulators should help teams raise quality, reuse what works and stop funding what doesn’t.
Three moves to start
- See the economics. Build one cross-provider view of AI accounts, API keys, tools and infrastructure, and attribute every cost to an owner, workload and measurable outcome.
- Steer the workload. Route each request by quality, cost, latency, resilience and data-control needs, with budgets and guardrails set per workload, not just per provider account.
- Prove and improve the result. Keep an audit trail of material activity and human approvals, measure quality beside unit cost, and turn the best-performing patterns into enterprise standards.
Where Cognizant AI Nexus fits
Cognizant AI Nexus was designed for this challenge: a modular, multi-model operating layer around whichever models a bank already uses. Banks can start with their most pressing constraint and add modules as they mature.
- Reveal maps tokens, tools and dollars to people, projects and outcomes.
- Route directs each request to the model with the right balance of quality, cost and control.
- Coach evaluates AI work in context and turns better practices into shared standards.
- Engage gives engineers and business users one governed workspace.
- Orchestrate converts business processes into audited agentic workflows, with human approval retained for critical decisions.
The aim isn’t another AI tool. It’s a governed environment where successful work is measured, reused and improved across teams.
Model capability will still matter, but models will keep changing. Durable advantage will come from how fast a bank can turn AI spend into outcomes it can explain, govern and repeat.
For CIOs, CFOs, CROs and COOs across ASEAN banking: if the board asked tomorrow what your AI investment returned, who owned the result and which controls applied, how long would it take you to answer?
Featured image credit: Edited by Fintech News Singapore, based on image by byzain via Magnific
