Agentic AI in banks won’t make workers obsolete – they’ll become more important - Opinion
Source: Business Times
Article Date: 11 Jun 2026
Author: Andrew Murray
As financial institutions implement the technology, firms that recognise the value of human judgment will be at an advantage.
As thousands of artificial intelligence builders, investors and policymakers gathered in Singapore on Wednesday (Jun 10) for SuperAI – a two-day conference for the frontier technology – the conversation is moving beyond AI-capability demos and productivity claims.
The real focus is on commercialisation: how companies can turn AI-driven efficiency into revenue growth, competitive advantage and new operating models.
Banking may become one of the clearest tests of that transition. The outcome will hinge on two things: the infrastructure that banks put under agentic AI, and the human judgment they build around it.
For years, banks have invested heavily in automation, analytics and AI assistants designed to make existing processes faster.
Agentic AI represents something more consequential. These systems initiate tasks, reason across data sources and execute workflows with limited human intervention. The implication is not simply lower costs. It is also a fundamental redesign of how banking work gets done and where human discernment creates value.
Consider a financial crime investigator. Until recently, much of their day involved gathering information from disconnected systems, compiling evidence and preparing reports before analysis began.
In a workflow built around agentic AI, that operational groundwork can arrive already completed. The system has assembled the transaction history, connected the related entities, mapped activity against known criminal typologies and presented the reasoning behind its conclusions.
The investigator’s role changes from collecting information to interrogating it. Their expertise becomes more concentrated, not less.
That shift matters because the commercial upside of agentic AI will not come from replacing judgment. It will come from allowing institutions to apply skilled assessment at greater scale and speed.
Rethinking AI infrastructure
The previous generation of AI in banking was largely assistive: fraud models flagged anomalies; rules filtered alerts.
Agentic AI changes the operating model itself. An anti-money laundering agent does not simply identify suspicious activity. It can also open a case, assemble supporting evidence, evaluate risk indicators and generate an initial recommendation for human review.
Similar architectures are emerging across onboarding, credit assessment and customer retention.
The commercial opportunity is obvious: faster decisions, lower operational friction and more scalable service models.
But the constraint is rarely the model itself; it is the architecture underneath. Many financial institutions still operate across fragmented environments built for an earlier era. Most were never designed for real-time AI orchestration across the full banking stack.
That fragmentation is a strategic problem. The institutions most likely to pull ahead may not have the most advanced AI models. They will be the ones capable of integrating data, governance, infrastructure and AI reasoning into a coherent operating environment.
This becomes even more important in regulated industries because large language models remain non-deterministic. That is, the same prompt can generate different outputs at different times. In consumer applications, that variability may be acceptable. In banking, it becomes harder to defend.
A suspicious-transaction report may be reviewed years later by regulators or law enforcement. A credit decision may be challenged by a customer. Banks need reproducibility, auditability and traceability. Without those safeguards, agentic AI struggles to move beyond experimentation into core operational workflows.
That is why explainability is becoming commercially important, not just a regulatory necessity. The Financial Crimes AI Agent announced in May by FIS and Anthropic was built around that principle: Every conclusion links back to its source data, and every decision stays with the investigator.
Singapore is positioning itself unusually well for this transition. The Model AI Governance Framework for Agentic AI launched earlier this year by the Infocomm Media Development Authority was the first of its kind globally.
The Monetary Authority of Singapore (MAS) followed with its AI Risk Management Toolkit under Project MindForge. MAS then went further. In partnership with the Government Technology Agency of Singapore and the Singapore Police Force, it used pooled transaction data from five banks to test AI scam detection.
That progression matters: The Republic is not approaching agentic AI as a theoretical policy discussion. It is accelerating practical deployment while shaping the rules in parallel.
But the institutions that capture the commercial upside will need more than the right rails. They will also need to invest just as much in their talent and upskilling.
Getting a leg-up on the competition
Discussions around the impact of agentic AI on the workforce often focus on what gets automated away. The more significant shift, and the one most likely to determine commercial outcomes, is how roles and responsibilities evolve.
As operational tasks become increasingly automated, human expertise concentrates in where it adds the most value. The investigator who once spent hours gathering information will spend more time evaluating outputs, challenging assumptions and identifying when technically plausible conclusions are operationally flawed.
That demands a different kind of professional development. Institutions will need people who can interrogate AI reasoning, not just consume AI outputs.
The technology itself has to support that process. An agent that exposes its reasoning can be challenged. One that simply delivers conclusions encourages overreliance.
Automation bias, where people gradually stop questioning machine outputs, is already documented in other industries. Financial services are unlikely to be immune.
Ironically, the more effective AI systems become, the greater that risk may grow.
The banks that benefit most from agentic AI will treat critical engagement as a competitive capability, not just a compliance requirement.
The organisations that convert AI productivity into sustainable revenue growth will not just automate workflows. They will redesign workflows around faster, more scalable human judgment.
The writer is head of international banking and payments, FIS
The commentary is based on the writer’s own experiences, observations and argument. AI tools were used for drafting. The writer remains fully accountable for the commentary’s accuracy, originality and final form.
Source: The Business Times © SPH Media Limited. Permission required for reproduction.
146