The Shift: What Happened
As artificial intelligence moves out of pilot labs and into everyday banking operations, a blunt reality check has arrived. In a commentary published on 20 November 2026 by Naveen Bankal, Head of Banking and Financial Services for ASEAN at Cognizant, the numbers show that most financial institutions still cannot turn AI spending into predictable, defensible results. The 2026 State of Tokenomics study of 472 organisations found that three in four could not confidently prove AI business outcomes to their CFO, while a Cognizant survey of 250 FinOps leaders showed 72% exceeded their AI budgets over the past year. Here in Malaysia, the gap is even sharper: an IDC study for Expereo reported that 80% of Malaysian respondents experienced cost overruns or missed ROI, compared with 54% across Asia-Pacific.
The issue is rarely the headline price of an AI token. Behind every customer-facing tool sits a hidden stack—vector databases, embeddings, orchestration, observability, and data movement—that can account for 40% to 60% of a feature's total cost. Worse still, 97% of organisations in Cognizant's FinOps survey had an autonomous "agentic AI" session escalate beyond expected cost or activity. Without proper tracking of cost per onboarding file processed, servicing request resolved, or alert reviewed—including failed runs in the math—banks end up paying frontier-model prices (50 to 100 times higher per token than smaller models) for routine work.
To fix this, banks across ASEAN are shifting toward governed, multi-model operating layers and dynamic model routing—a practice already used or evaluated by 86% of respondents, making them four times more likely to demonstrate real value. At the same time, regional regulation is tightening. On 7 October 2026, the Monetary Authority of Singapore issued AI risk management guidelines mandating strict AI inventories, risk-materiality assessments, and third-party oversight, with governance expectations starting October 2027, lifecycle controls in October 2028, and a consultation on agentic AI slated for 2027. For Malaysian banks operating regionally or competing for trust under BNM oversight, "build first, govern later" is no longer viable.
Why This Matters for Middle-Class Malaysian Families
When you swipe your credit card, apply for a housing loan, or check your unit trust balance on a banking app, you might wonder why back-office AI budgets matter to your household Ringgit. The answer comes down to two things every working Malaysian family relies on: the fees you pay and the safety of your data. When 80% of surveyed Malaysian organisations face AI cost overruns, those unchecked operational expenses don't just vanish—over time, banks under margin pressure look for ways to protect their bottom line through service charges, spreads, or slower customer-service upgrades.
Equally important is resilience and privacy when your salary, mortgage, and savings sit inside the banking system. As Naveen Bankal highlighted on 20 November 2026, proper AI routing isn't only about saving money; it keeps sensitive workloads inside the enterprise and ensures policy travels with every single request. Whether you are building an emergency fund, contributing consistently to EPF/KWSP and PRS for retirement, or paying premiums to protect your family's medical safety net, you depend on banks and financial platforms that verify every automated decision, keep an audit trail, and prevent runaway automated processes from mishandling your account or alerting you too late.
Strategic Action Steps
Just as banks must measure outcomes rather than raw activity to stop budget leaks, Malaysian households can apply the exact same "visibility and control" discipline to their own Ringgit this month:
- Audit your "hidden stack" of recurring fees: Just as vector databases and orchestration make up 40% to 60% of a bank's AI bill, small recurring bank charges, unused app subscriptions, and unmonitored card fees quietly drain household cash flow. Review your latest monthly statements and cut any service that delivers no measurable value.
- Right-size your financial tools: Banks shouldn't pay 50 to 100 times more for a frontier AI model when a leaner one handles the job. Apply the same filter to your money—check that your everyday savings, credit cards, and investment platforms aren't charging premium fees for basic transactions you can do at lower cost.
- Verify automated alerts and account controls: With banks increasingly relying on AI to review alerts and process servicing requests, log into your banking apps and double-check that transaction notifications, transfer limits, and fraud alerts are active and accurate so nothing slips past unattended.
- Keep your core safety nets separated and governed: Never leave your family's protection to guesswork. Confirm that your emergency cash reserve, medical coverage, SOCSO/PERKESO status, and long-term EPF/KWSP or PRS allocations are clearly documented—especially before year-end LHDN tax planning considerations come around.
Long-Term Wealth Perspective
Working with families across Kuala Lumpur, I often notice that financial stress rarely comes from one dramatic mistake; it comes from unmeasured spending and systems without guardrails. The lesson from the 20 November 2026 banking findings applies just as cleanly at the kitchen table as it does in a bank boardroom: speed and flashy new tools mean very little if you cannot explain the outcome, control the cost, and protect what matters.
Stay disciplined with your cash flow, demand transparency from the financial institutions holding your Ringgit, and keep your family's core assets—your emergency buffer, medical protection, and retirement nest egg—governed by clear rules rather than impulse.
Disclaimer: This article is provided for informational and educational purposes only and does not constitute formal financial, investment, or legal advice. Policy values and guidelines are subject to regulatory updates.
