AUG 27 — The next time an AI agent books a ride, selects a product or presses “pay”, the most important financial decision may occur before the transaction begins: deciding what the machine is allowed to do with our money.

Malaysia has already entered this new phase. In March 2026, the country saw its first authenticated agentic transaction pilot involving Mastercard, CIMB and RHB. In the pilot, an AI agent booked a ride from KLIA to KL Sentral on a user’s behalf, with consent and authentication built in.

This is more than a faster checkout. It marks a shift from technology that helps us make financial decisions to technology that can make and execute parts of those decisions for us.

An efficient transaction is not necessarily the best decision for a household. A lower instalment may still strain next month’s cash flow, a “best deal” may encourage an unnecessary purchase, and an automated renewal may preserve an expense the household no longer values.

AI can optimise a transaction without necessarily optimising financial wellbeing.

That is why financial literacy must evolve. Consumers will need to understand not only whether they can afford a purchase, but also how to define their priorities, set limits and decide which parts of their financial judgement can safely be delegated to an AI system.

Traditional financial education assumes that the person learning is also the person deciding. We teach consumers to budget, compare, borrow and invest because they are expected to make the final call.

But what happens when some of that judgement is delegated to a machine?

The issue is not whether AI can complete a transaction efficiently. It can. The harder question is whether efficiency always produces a sound financial decision.

That distinction matters as Malaysians become more comfortable using AI for financial decisions.

An HSBC-commissioned survey found that 85 per cent of affluent and high-net-worth respondents in Malaysia use AI tools for finance, placing Malaysia among the highest-adoption markets surveyed. Yet 58 per cent still preferred a combination of AI and human judgement when making financial decisions.

That caution is worth preserving. As AI moves from recommending to acting, consumers will need a new form of financial literacy: knowing not only how to make decisions, but also what should be delegated.

Three questions are critical. First, what exactly am I authorising? Giving an AI agent permission to “find the best option” is not the same as defining “best”. Does it mean the lowest price, lowest risk, best value or best fit with my budget? The quality of an automated decision depends partly on the goal we give the machine.

Second, what limits have I set? Consumers should define spending ceilings, time periods, product categories and situations requiring human approval. Convenience should not become open-ended authority.

Third, what happens if the agent is wrong? This goes beyond cybersecurity. OECD research notes that systems acting autonomously for consumers can expand agency but may also constrain autonomy, depending on their design, governance and market environment.

Consumers must therefore know whether they can challenge or reverse a transaction, and who was authorised to do what. Financial institutions and technology companies also have responsibilities.

Agentic commerce should provide meaningful consent, transparent limits, clear records of an agent’s authority and simple ways to pause or revoke it.

The payment industry is already moving in this direction. Mastercard’s Malaysian pilot emphasised consent, authentication and consumer control, while Visa is expanding programmes to prepare banks and payment providers for agent-initiated commerce across Asia Pacific.

Financial education must catch up. We have spent years teaching people to think before they spend. The next task is teaching them when technology may think and act on their behalf and when it must not.

AI will make finance faster, more personalised and more convenient. But progress should not be measured by how many financial decisions we automate. It should be measured by how much human judgement we preserve when the consequences remain ours to bear.

The next financial literacy test is not whether AI can be trusted with our money. It is whether we know how much of our financial judgement we are willing to give away.

*The author is D. Amirah Shazana Magli, who is a Senior Lecturer at the Faculty of Business and Economics, Universiti Malaya, and can be reached at amirahshazana@um.edu.my.

** This is the personal opinion of the writer or publication and does not necessarily represent the views of Malay Mail.