“People have been hearing all sorts of things about computers during the past ten years through the media. Supposedly, computers have been controlling various aspects of their lives. Yet in spite of that, most adults have no idea of what a computer really is, of what it can or can’t do.”

Steve Jobs said this decades ago, captured in Make Something Wonderful, a book of his own words. Sure, he was talking about computers back then.

But read it again today, and the misunderstanding he described hasn’t gone anywhere. It’s just wearing a different name.

Swap out the word computers for agentic AI, and you have a near-perfect portrait of where fintech discourse sits right now. Autonomous systems that go beyond answering questions to take actions, make decisions and execute end-to-end tasks with little to no human intervention.

Agentic AI talk is everywhere, and the expectations for it are enormous.

But underneath it all, the same problem Jobs identified persists: whether you’re engaging with agentic AI, building it, buying it, or regulating it, the real and present question is whether anyone truly understands what it is, what it can do, and where it breaks.

With agentic AI, the stakes of not knowing are categorically different. Systems no longer produce outputs alone; they also initiate actions. The shift from passive to active is precisely where the exposure begins.

In financial services and in fintech, that exposure has a name. When agentic AI in fintech is embedded into credit decisions, forex comparisons, wealth recommendations and customer experiences, it becomes a risk vector, touching credit risk models, compliance frameworks, customer outcomes and institutional reputation—all at once.

The industry is moving fast towards AI-first and AI-native operations. The harder question is whether clarity is keeping pace.

How Singapore’s Banks Turn Agentic AI From Hype to Value

Gartner predicts that over 40% of agentic AI projects will be cancelled by the end of 2027. Most are being driven by hype into early-stage experiments and proof of concepts that were never fully grounded in clear operational intent to begin with.

Institutions that succeed in extracting real value from this technology are going to be the ones that stop looking for a shortcut. They will start building the foundation with human oversight designed in, and where autonomy never replaces accountability.

In Southeast Asia, Singapore offers one of the clearest views of how financial institutions are attempting to close in on clarity.

Bank of Singapore, for instance, deployed an agentic AI tool called the Source of Wealth Assistant (SOWA). The tool automates an integral part of the KYC due diligence process, ensuring the legitimacy of clients’ wealth and transactions.

KYC for high-net-worth clients requires establishing the legitimacy of a client’s wealth and transactions against a dense body of regulatory expectations.

SOWA automates the core of the process, cutting the time it takes for relationship managers to produce a Source of Wealth report from 10 days to an hour, while still ensuring these align with regulatory standards.

Relationship managers review and refine the AI-generated draft before it moves to internal review teams for anti-money laundering and counter-terrorism financing assessments. The SOWA-processed data remains hosted on the bank’s private cloud.

Kam Chin Wong, Global Head of Financial Crime Compliance, Bank of Singapore, has said:

“With AI integrated into the source of wealth reporting process, relationship managers can shift their focus from manual documentation to meaningful client engagement and risk assessment. This not only strengthens client relationships but also maintains high standards of regulatory compliance while delivering greater value.”

In a broader context, OCBC has taken that same philosophy and embedded it across how it touches the bank’s operations. Over six million decisions are AI-powered daily, spanning revenue growth, risk mitigation and productivity. Every in-house tool is built against the FEAT principles of Fairness, Ethics, Accountability and Transparency, with regular reviews to test for accuracy and screen for bias across gender, nationality and other dimensions.

DBS, meanwhile, has pushed into newer and more consequential territory as the first bank in the Asia Pacific to pilot AI-powered agent payments via Visa’s Intelligent Commerce. The pilot actively tests how agent-initiated transactions can move through existing card network infrastructure under issuer-controlled, secure processes.

The exercise will assess how AI-driven transactions can be integrated into existing systems while maintaining regulatory, operational and security standards. The bank is simultaneously stress-testing the authentication architecture that agent-led payments will depend on, with controls sitting at both the issuer and network level.

T.R. Ramachandran, Head of Products & Solutions, Asia Pacific at Visa, shared,

“Through Visa Intelligent Commerce and Trusted Agent Protocol, we’re building the foundation that will make agentic commerce safe, secure and scalable — from AI‑ready credentials to advanced authentication. This sets the stage for how trusted, AI‑powered experiences will come to life for consumers and partners across the region.”

By January 2026, DBS reported that its AI initiatives had generated S$1 billion in economic value in 2025, compared to S$750 million the year prior, a figure derived from comparing the outcomes between AI-enabled customers and control groups.

The common thread across these use cases is clarity, applied comprehensively around agentic AI deployment.

Closing the Clarity Gap

If the history of computing has taught us anything, it is the fact that the most powerful tools are also the ones that are most prone to being misunderstood. As agentic AI moves from generating text to executing financial tasks, the “understanding gap” scenario Steve Jobs identified decades ago resurfaces and compounds.

Every layer of autonomy added without sufficient comprehension is another layer of exposure accumulating until something makes it a visible problem.

To bridge this gap, financial institutions must stop looking for a quick fix and build a foundation that allows for human-in-the-loop oversight, ensuring autonomy never outpaces accountability. The differentiator is clarity on the end goal.

In fintech, the future will be shaped by businesses that master the discipline of knowing precisely when to keep humans in command.

If you want to understand more about how Southeast Asia’s leading banks and fintechs are operationalising agentic AI, watch the full webinar on Beyond the Bot: Agentic AI’s Evolving Role in Financial Services.