A year or two ago, bank executives used to ask how artificial intelligence (AI) could make employees faster or shave costs off operations.

Garry Sien, Chief Innovation and Solutions Officer, International Business at Ant Digital Technologies, says that conversation has moved on.

“Since the beginning of 2026, that interest level has now swung into revenue generation,” he told Fintech News Network. Bank executives put it more bluntly now. Show me the ROI, they say.

The shift matters most in markets still building out formal banking access.

In the Philippines, the share of adults with a formal financial account climbed to 58% in the first quarter of 2026, up from 48% a year earlier, according to a survey commissioned by the Bangko Sentral ng Pilipinas (BSP), with e-wallets driving most of the growth.

Data Quality, Not More AI, Is the Real Barrier

Getting AI from pilot to production remains the bigger obstacle, and it usually is not the model itself.

“The first barrier that they encounter often is not the AI itself; it’s actually the quality of their data and their knowledge,” Garry said.

Institutions with messy data need larger, costlier models to compensate.

Ant Digital works with banks to structure their data first, before deciding which models suit which task, since a well-organised bank can often use smaller, cheaper models.

Garry said the hardest part was not technology integration but getting business teams, not just the IT department, to own the project.

Writing code itself is no longer the constraint.

Some of Ant’s internal teams now generate around 90% of their code with AI, Garry acknowledged. He believes that what matters most now is knowing what to build.

GCash’s Onboarding Overhaul

GCash offers a more familiar example in Southeast Asia.

Before working with Ant Digital, Garry said GCash was already onboarding customers remotely.

Identity checks, however, could involve a human agent calling the applicant by video to verify who they were.

Even when the process went smoothly, completing it could take days.

GCash came to Ant looking for a faster way to handle remote identity verification without weakening the checks needed to satisfy regulatory and risk requirements.

ZOLOZ, which began as the KYC technology used by Alipay, was subsequently adapted to support GCash’s eKYC journey.

The Philippines presented its own complication.

Unlike markets where customers rely on only a handful of commonly accepted identity documents, Garry mentioned that Filipinos can use a much wider range of IDs when opening accounts.

Ant’s teams therefore had to train the system to recognise those variations while combining document checks with biometric recognition.

Digital onboarding has since developed another challenge, which is knowing whether the face appearing on screen belongs to a real person, is becoming considerably harder.

Garry explained that GCash upgraded its ZOLOZ deployment in 2026, adding more advanced anti-deepfake capabilities alongside improvements designed to let more legitimate applicants complete onboarding successfully.

Thrift and Rural Banks See An Opening

Digital banks with newer core systems can apply AI more directly than institutions carrying decades of legacy infrastructure, Garry said, pointing to Malaysia’s Ryt Bank, which has handled double-digit concurrent account openings per second during acquisition campaigns.

Traditional banks must instead retrofit AI without breaking existing systems and controls.

Smaller, established institutions face the same retrofit problem, but Garry argues size does not have to be a disadvantage.

Fewer branches and smaller balance sheets can also mean fewer systems to untangle, letting smaller banks move faster once they commit.

The Philippines’ own thrift banks are testing that theory.

Net profit for the sector rose 11.8% to PHP 5.86 billion in the first quarter of 2026, with fee income more than doubling, even as it faces tougher competition from digital banks and fintechs, according to the Chamber of Thrift Banks (CTB).

CTB president Jaime Valentin Araneta, who also runs Equicom Savings Bank, told Philstar that thrift banks are using AI to cut manual back-office work and free up staff for higher-value tasks.

The same tools are improving credit decisions and fraud detection.

“Digital transformation is an existential imperative for banks,” Araneta said.

Though he cautioned that the fraud-monitoring systems now required by law can be costly, particularly for banks relying on third-party providers.

Who is Accountable When AI Acts On Its Own?

The stakes rise further once AI stops recommending and starts acting.

Agentic AI could eventually carry out tasks on a customer’s behalf, raising questions of trust and accountability that Garry frames in unusually human terms.

Earlier technology still needed a human operating it directly.

An AI agent instead functions as an extension of its user, right up until it misreads an instruction involving someone’s money.

“You’re still accountable,” Garry answered

He feels like we should be treating AI more as an extension of the person using it rather than a separate actor.

But not every institution has adjusted how it works to match the investment going into AI, Garry pointed out.

Dropping AI into an unchanged organisation, without adjusting the processes, controls and audit practices around it, tends to disappoint.

Araneta made a similar case to his own members, warning that banks which skip on cybersecurity spending now will end up paying far more once a breach forces the issue.

From Single AI Tool to Agent Factory

Banks are already asking harder questions about what they get back from AI spending, and Garry thinks that will affect the kind of technology they buy next.

Instead of adding another standalone feature, some are looking for systems that can take on several jobs across the business.

Microsoft has been pursuing a similar idea through its “Agent factory” strategy, although its approach differs from Ant Digital Technologies’.

“Customers want a whole solution,” Garry said.

Ant Digital Technologies introduced what it calls the “Commercial Agent Super Factory” at WAIC 2026.

The idea is to give enterprises a way to deploy multiple agents for different roles rather than rely on one tool for one use case.

Banking is a natural fit for that model. Underwriting may need customer data and risk checks, while KYC and fraud detection rely on their own systems, rules and workflows.

One AI feature is unlikely to cover all of it.

When Ant Digital Technologies introduced Agentar 2.0 at WAIC 2026, the platform included nearly 200 role-based digital expert templates and hundreds of skill modules that institutions could subscribe to and combine based on what they needed.

Garry believes the model could be particularly useful for banks that want to expand their use of AI without building every capability themselves, especially as pressure grows to show a return on those investments.

Ant Digital Technologies is now carrying that approach into its overseas strategy.

Featured image: Edited by Fintech News Singapore based on an image by Ant Digital Technologies.