Payment Gateways

Taiwan: Neweb Launches AI Shopping Agent With Mastercard Agent Pay, Starting With Travel

Taiwanese payment gateway Neweb has launched BuyTalk Agent, a platform that lets AI assistants recommend products, place orders and pay for them on a shopper’s behalf. Built with local partner Yingyuan Technology and plugged into Mastercard’s Agent Pay, the service starts with travel merchants and could later reach the 400,000-plus businesses that already use Neweb to collect payments.

Key facts

  • Neweb (known in Chinese as 藍新科技, part of Taiwan-listed Soft-World International) announced BuyTalk Agent on 12 November 2026.
  • The platform was co-developed with Yingyuan Technology and integrates Mastercard Agent Pay to handle payments made by AI agents.
  • Neweb says it works with more than 400,000 partner merchants.
  • The first phase targets the travel industry; Neweb plans to extend it to other sectors later.
  • A Mastercard survey cited at the launch found 57% of Taiwanese respondents would let an AI agent find products within their preferences and budget and buy once they confirm.

How BuyTalk Agent works

The idea behind agentic commerce is simple: instead of a shopper clicking through websites, an AI assistant does the searching, comparing and checking out. The hard part is making sure the assistant only buys what the person really wants, and that card details are not exposed along the way.

Neweb splits the process into four steps. First, a merchant’s product information is turned into structured data that AI systems can read. Second, the platform matches that catalogue to what the AI has understood about a shopper’s needs. Third, the order is confirmed. Fourth, the payment is authorised.

Merchants do not need to build their own AI models. They only supply product details, and the platform converts them. Neweb says the same structured data could later be used by other large language model (LLM) services, which may give smaller sellers a way to appear in AI-driven shopping without heavy technical work.

On the payment side, Mastercard Agent Pay brings two security tools. Network tokenisation replaces the real card number with a substitute code, so the actual number is not shared with the AI agent or merchant. Payment passkeys let the cardholder confirm the purchase with a fingerprint, face scan or device PIN rather than a password or one-time SMS code.

Neweb general manager Chung Hsing-po described AI as having moved beyond a support tool to become a new bridge between merchants and consumers.

Why it matters

Taiwan’s online shoppers appear ready to try the idea. Besides the 57% willing to let an agent complete a purchase, the Mastercard survey found more than 60% would let AI plan a travel itinerary and handle the payment. That helps explain why Neweb chose travel, where bookings involve many choices on dates, prices and add-ons, as its first market.

For the region, the launch shows that agentic payments are moving from global card-network announcements to local payment gateways that actually serve merchants. A gateway sits between shops and the card networks and banks, so when one with hundreds of thousands of merchants adds agent-ready checkout, the technology can reach ordinary businesses much faster than if each shop had to build it alone.

It also raises questions that regulators across Asia are starting to ask: who is responsible if an agent buys the wrong thing, how consent is recorded, and how disputes and refunds should work when no human pressed the final button. Tools such as tokenisation and passkeys address security, but rules on consent and liability are still being written in most markets.

What to watch

  • When BuyTalk Agent goes live for travel merchants, and which sectors Neweb adds next.
  • Whether other Taiwanese gateways and banks join Mastercard Agent Pay or rival schemes from Visa and others.
  • How Taiwan’s financial regulators approach consent, liability and refunds for purchases made by AI agents.


Related on FinDeskAsia: Hong Kong: Deutsche Bank and Ant Digital set rules for AI agents that move company money and how payment gateways work.

Sources: MoneyDJ (November 2026); Leho (November 2026); Commercial Times (November 2026).

FinDeskAsia

FinDeskAsia editorial desk — finance made clear across Asia.