Advertisement

Beyond the AI Hype: Is Bangladesh Ready for Agentic Banking?

Beyond the AI Hype: Is Bangladesh Ready for Agentic Banking?
Md Mahmudul Hasan. TIMES Sketch
Advertisement
Advertisement
Advertisement
Advertisement

Three days at the Singapore FinTech Festival 2026 left me with one overwhelming impression: artificial intelligence is no longer a side exhibit. It is the exhibition.

AI-powered e-KYC, intelligent customer service, instant lending engines, behavioural credit scoring and voice agents that sound uncomfortably human all point in the same direction. Language models are moving from experimentation to real-world deployment. There is hype, certainly. But beneath it lies a structural shift that Bangladesh cannot afford to misread.

From automation to agency

Traditional automation follows predefined instructions. Cognitive AI can interpret information, reason and adapt. Agentic AI goes a step further. It pursues goals, coordinates tasks across systems and executes authorised actions on behalf of customers. Imagine an AI agent checking loan eligibility, preparing an application, comparing financial products and scheduling repayments around a customer’s income.

At the launch of his new book, Bank 5.0, Brett King argued that AI agents could become the dominant users of banking services, with the primary financial relationship increasingly shifting towards a customer’s personal AI agent rather than any single bank. The implications extend well beyond banking.

Capital has already placed its bet. On 19 November, Anthropic, the company behind Claude, began trading on Nasdaq following an IPO priced at $83 per share. With the underwriters’ option exercised in full, the offering generated approximately $93.5 billion in gross proceeds, while a concurrent $26 billion private placement brought the combined total to approximately $119.5 billion. The company’s fully diluted valuation was around $1.5 trillion. These figures reflect extraordinary investor confidence in AI’s commercial potential. But they also raise difficult questions about infrastructure costs, profitability and sustainable business value. Our banks should take note: investment in AI must be matched by a clear business case.

The dark side of agency

Criminals are adopting the same technologies. AI can lower the barriers to sophisticated cyberattacks, accelerate reconnaissance and make social engineering more convincing. In 2025, Anthropic’s threat intelligence team documented an AI-orchestrated cyber-espionage campaign. In July 2026, researchers at Israeli security firm Dream reported a near-autonomous campaign against Taiwanese government agencies that deployed as many as eight AI agents simultaneously. Google’s Threat Intelligence Group and Mandiant have also warned that adversaries are moving towards agentic operations.

Financial fraud is evolving alongside these threats. In Hong Kong, police working with mainland authorities uncovered a syndicate that used pre-recorded footage to bypass digital banks’ facial recognition systems, laundering more than HK$270 million (US$34.7 million). In the UK, authorised push payment fraud losses rose 19 per cent to £576.4 million in 2025. Deloitte projects that generative AI could push US fraud losses to $40 billion by 2027, up from $12.3 billion in 2023.

Advertisement
Advertisement

Bangladesh is not insulated. According to its FY25 annual report, the Bangladesh Financial Intelligence Unit (BFIU) received a record 30,199 suspicious transaction and activity reports, an increase of 74 per cent, with nearly 95 per cent originating from banks. In June, Finance Minister Amir Khosru Mahmud Chowdhury told parliament that the BFIU had suspended or frozen around 55,000 mobile financial services accounts over alleged links to online gambling and digital hundi. A Policy Research Institute survey published in 2022 found that nearly one in ten MFS users had experienced fraud.

As deepfakes, synthetic identities and AI-assisted fraud become more accessible, a selfie and a scanned National Identity Card can no longer be treated as sufficient proof of identity. Financial institutions must rethink identity verification, behavioural intelligence and transaction monitoring as interconnected capabilities rather than isolated controls.

Network intelligence, not isolated alerts

Singapore’s approach offers an important lesson. Effective financial crime prevention increasingly depends on analysing relationships between accounts, devices, beneficiaries and transaction patterns to identify suspicious networks that may remain invisible within any single institution. This matters equally for fraud prevention, anti-money laundering and counter-terrorist financing.

At the festival, Singapore’s Deputy Prime Minister and Monetary Authority of Singapore (MAS) Chairman Gan Kim Yong announced three National AI Missions for financial services: strengthening financial crime prevention, improving access to financing and enabling safe agentic transactions.

Through an AI Innovation Arena, MAS and the Future of Finance Institute will give developers a synthetic dataset reflecting interbank transaction patterns and scam activity pooled from five banks. The Institute will also establish an Agentic Money Sandbox, where financial institutions can test how AI agents transact with digital money within defined authority, limits and safeguards.

Read more

Separately, MAS’s Safeguards for Agentic Finance at Runtime (SAFR) framework, published in July, provides an important foundation for governing AI agents in financial services. The next phase of Singapore’s work will explore common approaches to agent identity, authority, accountability and interoperability.

This points towards an emerging concept: Know Your Agent. Just as banks verify customers, they will increasingly need to establish whether an AI agent is legitimate, whom it represents and what it is authorised to do.

This is precisely the direction Bangladesh needs to consider. Financial crime does not respect institutional boundaries, and neither can its prevention.

Where Bangladesh stands

I remain cautiously optimistic, with the emphasis on caution. Bangladesh Bank’s Cybersecurity Framework, Version 1.0 (2026), requires banks, finance companies, MFS providers and payment operators to ensure compliance by 31 December 2026. The central bank has also initiated work towards an AI-based cybersecurity policy. These are important foundations.

However, our biggest obstacle is not technology. It is mindset. Fragmented data, legacy infrastructure and an organisational culture that often treats change as a risk rather than an opportunity could lead us to bolt AI onto broken processes, automate yesterday’s inefficiencies and call it transformation.

Yet the opportunity is substantial. Intelligent financial services could help expand credit access for small businesses, improve financial products for underserved communities, reduce operating costs and deliver more personalised services. For a country where financial inclusion remains a development priority, AI should do more than make existing services faster; it should make financial services more accessible, affordable and responsive.

Five priorities for Bangladesh

First, treat data as critical infrastructure: properly governed, protected, consent-based and interoperable. Without trusted data, even the most advanced AI models will produce unreliable outcomes.

Second, build a national, privacy-preserving fraud and mule-account intelligence network connecting banks, MFS providers, BFIU and law enforcement. Financial crime intelligence must move beyond institutional silos.

Third, move beyond KYC towards Know Your Agent, establishing how an AI agent is authenticated, whom it represents, what authority it holds and where its transaction limits begin and end.

Fourth, establish a regulatory sandbox for agentic payments and AI-driven financial services before autonomous financial transactions become mainstream. Innovation must be tested in controlled environments, with clear accountability and consumer protection.

Fifth, invest in people. Financial institutions need professionals who understand technology, risk, regulation and business strategy. Technology can be purchased, but judgement cannot.

Above all, AI-driven decisions must remain explainable, accountable and subject to appropriate human oversight. Autonomy without safeguards is simply exposure.

The real opportunity

The real opportunity is not to introduce AI into business processes, but to redesign business processes around intelligence. Technology provides the infrastructure, cognitive capabilities provide the intelligence, and business strategy turns that intelligence into measurable value.

For Bangladesh, this is also an opportunity to rethink the architecture of financial services. Banks, MFS providers, fintech companies and technology platforms must move beyond isolated digital channels towards interoperable ecosystems where intelligence, trust and financial services work together.

Bangladesh leapfrogged into mobile money. We can leapfrog again, but this time the transformation must be built on trusted data, intelligent infrastructure, responsible innovation and a clear business vision.

The question is no longer whether AI will change banking, but whether our institutions are prepared to lead that change or simply react to it. Otherwise, we risk building digital banks for yesterday while the world moves towards intelligent banking.

The writer is a digital finance strategist and author of From Cash to Code: How Digital Finance Can Transform Bangladesh’s Invisible Economy and Visible Work, Invisible People: How Invisible Workers Power Bangladesh’s Economy.

Follow TIMES on Google News

Get trusted updates and editor-picked stories in your feed.

Follow
Related News