AI Goes Beyond Shopping to Payment

The era is approaching where AI will handle finding hotels, comparing prices, and even the payment process at a travel destination. With Macau's e-wallet MPay adopting Ant International's AI payment protocol, the trend of AI agents performing actual transactions beyond simple product recommendations is gaining momentum.

▲ Photo: AI-generated image

■AI moves to the payment screen

Macau's e-wallet MPay announced on September 11th its adoption of Ant International's 'Agentic Mobile Protocol (AMP)'. AMP is a protocol that allows AI agents to invoke mobile payment functions with the user's approval.

This works by the user telling the AI agent their travel itinerary and conditions, then the AI searches for hotels, compares prices and locations, and finally requests payment from the e-wallet for the user-selected product.

For example, if a user requests, “I need to attend an event in Shanghai, please book a hotel near the venue for one night,” the AI searches for accommodations and recommends suitable options. Once the user confirms the room type and approves the payment, MPay processes the payment. The user does not need to repeatedly switch between hotel websites and payment apps.

However, AI does not execute payments completely autonomously. According to MPay and AMP's explanations, users set the approval scope and payment conditions, and user authentication is required during the final payment process.

■From 'Recommendation AI' to 'Transaction AI'

Until now, the consumer's online purchase process generally followed this sequence:

Search → Compare → Decide → Pay

When an AI agent intervenes, this process changes as follows:

Goal Input → AI Search/Compare → Product Recommendation → User Approval → AI Payment Request

The core of this change is not that AI provides more information. It lies in AI interpreting user requirements, connecting various services, and executing actual transactions.

Meta's personal AI agent 'Muse', recently launched in the US, is also designed to perform multi-step tasks such as sending emails, booking travel, shopping, and managing schedules. Meta explained that it requests user approval before executing sensitive tasks and allows users to directly manage connected apps and permissions.

AI agents are no longer just tools that provide answers; they are becoming 'agents of action' that move through the digital world on behalf of users.

■AI Shopping, 300 Million Projected by 2030

While AI-driven shopping and payments are still in their early stages, related companies anticipate market expansion.

Mastercard recently predicted in a report that by 2030, over 300 million online shoppers worldwide could purchase and pay for goods via AI agents. This projection suggests that more than 1 in 10 online shoppers will use AI for purchases.

Mastercard expects AI agents to primarily handle repetitive and low-risk purchases such as groceries, medicines, household items, and subscriptions. Conversely, high-value products or financial instruments are likely to continue requiring direct user judgment and approval.

Differences in attitudes between adolescents and parent generations were also observed. A Mastercard survey of 26,000 parents and adolescents aged 13-18 across 13 markets found that 27% of adolescents were likely to use fully automated AI shopping assistants, higher than 16% of the parent generation. The response indicating trust in AI product recommendations over friends was also higher among adolescents.

However, it should be noted that this survey was commissioned by Mastercard, and the target countries were 13 European-centric markets. It is more appropriate to interpret this data as showing generational shifts in perception towards AI shopping, rather than definitive statistics on global consumer behavior.

■How will payment authority be managed?

Once AI begins to participate in payments, the most critical issue is not convenience but authority management. Users need to set the following limitations for AI agents:

-Amount that can be used at one time

-Categories of products that can be purchased

-Available merchants

-Conditions requiring mandatory approval before payment

-Scope for automatic recurring payments

-Personal information and accounts AI can access

MPay and AMP proposed a system that manages agent identity verification, user approval, and payment security as a single framework. Users can set the scope of authority and spending limits for each transaction, and modify or revoke approval details.

The purpose of this structure is to ensure that AI agents do not have unlimited authority, even when making payments on behalf of the user.

■The Role of Blockchain and Digital Wallets

If AI agents perform shopping and payments, the role of digital wallets may also change. If traditional wallets were places to store payment methods, wallets in the AI era will need to manage the following functions:

-Which AI agent requested the transaction

-What permissions the user granted

-What the payment limit is

-Under what conditions the transaction was approved

-Whether a transaction can be canceled or traced if a problem occurs

Blockchain has the potential to be used as a means to verify transaction records and AI agent permissions. This involves assigning a separate digital identity to each AI agent and recording user-approved conditions and transaction details in an unalterable form.

However, it is difficult to conclude that blockchain or virtual assets will be essential for AI payments. This is because, as seen in the MPay case, connecting existing e-wallets and card payment networks with AI agents is entirely feasible.

The significant change now is not so much in the payment method itself, but in redesigning authority and responsibility to allow AI to participate in the payment process.

■Scenes Likely to Appear First in Korea

In Korea, AI payments are likely to become a reality in the following areas of daily life:

Travel Booking

AI searches for flights and accommodations and proceeds with reservations according to the user's set budget and conditions.

Repetitive Shopping

AI can order regularly purchased items such as bottled water, pet food, and household goods based on price and delivery conditions.

Subscription Management

AI can compare telecommunication, OTT, and software subscriptions and suggest canceling unnecessary services or changing plans.

Overseas Payments

Travelers request local transportation, restaurant, and accommodation services from AI, and pay using existing e-wallets or simple payment methods.

For such services to spread domestically, not only must shopping, map, booking, and payment platforms be integrated, but also standards for personal information protection, electronic financial transactions, and consumer damage relief must be established.

The next stage of AI life is not more natural conversation. It is the automation of actions where AI understands user demands, finds products, compares conditions, and connects to payment.

MPay's AMP integration in Macau is significant because AI agent payments have moved beyond the laboratory and started connecting with actual e-wallets. However, the moment consumers grant payment authority to AI, they take on new responsibilities along with convenience.

The question in the era of AI agents is not “What does AI recommend?”

It is, “How far will I allow AI to conduct transactions on my behalf?”

Answering this question requires not just smarter AI, but also a clear approval system, a wallet with controllable limits, and verifiable transaction records.