AI is changing how we spend. Financial literacy must catch up — Amirah Shazana Magli

August 27 — When an AI agent initiates a financial transaction, the most crucial decision often takes place prior to the actual purchase. In Malaysia, this new era commenced in March 2026 with the country's inaugural authenticated agentic transaction trial between Mastercard, CIMB, and RHB. At the core of this shift lies the capability of AI to not only streamline processes but also execute parts of financial decisions on behalf of users.

This represents a departure from technology assisting in financial decisions to technology autonomously making and executing those decisions.

This development underscores the necessity for an evolution in financial literacy. Consumers must not only ascertain their ability to afford a purchase but also define their priorities, establish limits, and determine which aspects of their financial judgment can be entrusted to AI systems. While traditional financial education focused on budgeting, comparing, borrowing, and investing, this new context demands an additional layer of understanding: determining what aspects of financial decisions should be delegated to machines.

Despite Malaysia's high adoption rate of AI tools for finance, with 85% of affluent and high-net-worth individuals utilizing such tools, a significant majority (58%) still advocate for a blend of AI and human judgment in financial decision-making. This cautious approach is vital in navigating the emerging landscape where AI moves from merely recommending to actively acting on behalf of consumers.

Three key questions emerge as AI transitions from recommending to acting: Firstly, what precisely are you authorizing? Consent for an AI agent to seek the "best option" differs significantly from defining "best" – the criteria could range from the lowest price to the best value or the most financially aligned with your budget. Secondly, what limits have you set?

Consumers should delineate spending ceilings, timeframes, product categories, and situations necessitating human intervention. Convenience should not equate to limitless authority. Lastly, what happens if the AI agent errs? This extends beyond cybersecurity concerns, as highlighted by OECD research which notes that autonomous systems for consumers can enhance agency but may also impede autonomy based on their design, governance, and market environment.

Consumers must be aware of their ability to contest or reverse transactions, as well as the entities authorized to perform such actions.

Financial institutions and technology providers bear their own responsibilities. Agentic commerce should facilitate clear consent, transparent boundaries, explicit records of an agent's authority, and straightforward methods to pause or retract this authority. The payment industry is already progressing in this direction, as evidenced by Mastercard's Malaysian pilot, which emphasized consent, authentication, and consumer control.

Visa is also advancing by expanding programs to prepare banks and payment providers for agent-initiated commerce across the Asia Pacific region.

However, financial education must keep pace with these technological advancements. While years have been dedicated to instilling the habit of thoughtful spending, the forthcoming challenge lies in educating consumers about when technology may autonomously make decisions on their behalf and when human judgment remains paramount. The impending financial literacy test is not about the trustworthiness of AI with our finances; it is about our comprehension of how much of our financial judgment we are prepared to relinquish.

AI can undoubtedly expedite, personalize, and simplify financial transactions. Yet, the true measure of progress should not be the volume of automated decisions but the extent to which we preserve human judgment in circumstances where the implications still lie squarely on our shoulders.