Around 70 per cent of organisations in the Asia-Pacific (APAC) region expect agentic artificial intelligence (AI) to significantly disrupt their business models within the next 18 months, according to banking technology provider Moneythor.

The company said the banking sector is likely to see the greatest impact, with productivity and efficiency improvements from agentic AI projected to “exceed expectations” over the next three to five years.

‘Exponential’ potential and risks

Vivek Seetharaman, Chief Product Officer at Moneythor, described agentic AI’s influence on banks as “exponential” — both in terms of opportunities and risks.

“Agentic AI will help banks realise the true potential of what we call ‘Deep Banking’; personalised experiences that anticipate customer requirements, and extend beyond traditional banking services,” he said. “But make no mistake — so are the risks.”

Risks highlighted include data misclassification, such as an overdraft being misrepresented as a positive balance, and increased vulnerabilities from autonomous AI agents interacting across multiple proprietary and third-party systems.

Unlike traditional rule-based personalisation systems, agentic AI can “learn on the job”, refining its decisions over time. However, if the data is flawed, errors may compound rapidly, raising concerns for both banks and customers.

Guardrails and governance

Seetharaman emphasised that agentic AI deployment in banking cannot be treated as “plug and play”. He outlined three principles as essential to safe adoption:

Guardrails around risky language and contexts, applicable across different large language models (LLMs). Clear governance procedures for data usage, aligned with relevant jurisdictional regulations. Human oversight at regular intervals throughout AI decision-making processes.

Moneythor’s new AI Suite

The comments accompanied the launch of Moneythor’s new AI Suite for banks, which includes agentic AI alongside predictive, generative and conversational AI capabilities.

• Predictive AI: Forecasts cash flows and generates real-time contextual recommendations.
• Generative AI: Produces and adapts personalised customer content.
• Conversational AI: Uses Model Context Protocol (MCP) to deliver contextual dialogues across different LLMs without retraining.

Moneythor, which has been deploying AI in banking for over 13 years, counts DBS, Standard Chartered, Trust Bank and RHB Bank among its clients in Southeast Asia.

Agentic AI is expected to become increasingly central to customer relationships, supporting acquisition and service in ways that mimic human adaptability. However, experts stress that without careful governance, the risks may prove as significant as the rewards.