Invest1 publisher2 min readPublished
Ant International expects 95% of its merchants to onboard a payment method in an hour by end-2026
Ant International reports that merchants using AI to integrate its payment services quadrupled in a year, and that 89.5% of Antom merchants now run AI agents. Both figures are its own count. The one-hour onboarding is still a forecast for the end of 2026.
The Investor · Invest desk
What happened
- Ant International says the number of merchants using AI to integrate payment services with it quadrupled over the past year, as it extends AI into payments, foreign exchange, treasury and credit.
- The company expects 95% of merchants to be able to onboard new payment methods within one hour by the end of 2026.
- On Ant's own reporting, 89.5% of merchants using its Antom service have deployed AI agents, and AI assists with 81.4% of payment tasks.
- Nearly 100 products across the Alipay+, Antom, WorldFirst and Bettr businesses are scheduled to reach global markets through the second half of 2026.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- constraint The numbers are company-reported shares and multiples. An acquirer benchmarking its own onboarding roadmap against Ant has a growth rate and an adoption percentage to work with, and no volume figure to price them into.
- decision About 25 launches per brand in one half-year window commits Ant's engineering to breadth across four separate merchant bases, and defers depth in any single one.
- capability A credit decision returned in as little as a minute lets lending sit inside the payment flow, so a merchant can be underwritten while it is being onboarded.
Quadrupling is a growth rate. Ant did not publish the count it multiplies [1][3]. The two Antom figures are more useful divided than read side by side. 81.4 over 89.5 is about 91%. So if every AI-assisted payment task sits at a merchant that has deployed an agent, roughly nine tasks in ten inside that group already run through one [4][2].
Ant put numbers on the machinery. Antom Autopilot is a conversational interface across the merchant payment lifecycle, covering onboarding, integration, payment routing, fraud protection, revenue recovery, troubleshooting, reconciliation and market expansion [6]. Underneath it, the Antom 3-in-1 Transformer has more than 10 billion parameters, built to handle sequential, tabular and graph data in a single architecture [7]. Ant says it processes about 90 trillion data points a year [8]. That is about 247 billion a day [3].
The figure with cash attached is the hedging one. FalconTST, at more than 8.5 billion parameters, reached better than 93% accuracy in foreign-exchange and cash-flow forecasting in Ant's own real-world cases [9]. The company says those cases cut corporate hedging and allocation costs by 30% to 60% [10]. The top of that range is twice the bottom [6]. A treasurer sizing a switch on it budgets the 30 and treats the 60 as upside.
On the account side, WorldFirst is adding natural-language and agent-based workflows for account onboarding, payments and treasury. The Wyn assistant is aimed at SMEs, and the enterprise product gets supervised workflows for compliance, electronic know-your-customer and know-your-business checks [14]. In lending, Bettr's credit engine reaches more than 100 million users through 49 institutions in 28 markets [11], which averages about 2 million users per institution [4].
The standards work is what could bind other acquirers. Ant has been working with Mastercard and Visa on standards for identifying and governing AI agents in payment environments, building on its Alipay+ Agentic Mobile Protocol and know-your-agent controls [13]. If the card networks publish agent identification, an hour-long onboarding becomes something a merchant can ask of any acquirer, and Ant's adoption shares turn into a floor its rivals get measured against. If the controls stay inside Ant's own protocol, the percentages describe one channel's internal tooling.
I would weight the second case higher for now, because every quantified item here is Ant's own count [3] and the standards work is described as work in progress [13]. What would move me: a published specification with named issuers attached, or a merchant count under the quadrupling that lets the growth rate be converted into volume.
What to watch
- Whether the 95% figure arrives at end-2026 and Ant publishes a definition of what counts as onboarding a payment method.
- Whether FalconTST's 93% accuracy and the hedging saving get restated outside Ant's own selected cases.
- Whether Bettr's 49 institutions produce disclosed credit balances alongside the 100 million user reach.