AXS, Visa and Mastercard partner to let AI agents pay bills without multiple app logins
Sign up now: Get ST's newsletters delivered to your inbox
AXS’ upcoming service removes the need to manually check for outstanding bills, log into multiple apps and key in one-time passwords to make payments.
PHOTO: VISA, AXS
- AXS, Visa and Mastercard plan AI payments in Singapore for bills, loans and fines. The service links AI agents to AXS’s network of over 800 billers.
- Customers will control what the AI can do. Safeguards include payment limits, approved transactions, biometric checks and hidden tokens that protect card details.
- AXS will pilot the service with Grab AI assistant. Visa is also working on dispute handling and more bank checks as agentic payments are still new.
AI generated
SINGAPORE – Consumers may soon be able to pay their telco, electricity and healthcare bills, as well as housing loan instalments and fines, by speaking or typing a prompt to an artificial intelligence agent to do so on their behalf.
Global payments companies Mastercard and Visa are working with e-payment services provider AXS to roll out this service in Singapore by connecting AI agents – bots that can take actions with minimum human supervision – to AXS’ payment network of over 800 billing organisations.
These organisations include hospitals, power companies, telcos, insurance companies, finance companies, country clubs and government agencies such as the Housing Board and police.
AXS declined to specify the roll-out timeline for its agentic bill payment service to consumers. The firm is also planning a trial for bill payments with the Grab AI assistant.
“By enabling partners to connect their own AI agents to AXS’ bill payment network, we can extend our reach beyond our existing channels while empowering partners to create new payment experiences for their customers,” Jeffrey Goh, AXS’ group chief executive and executive director, said on Nov 18.
Existing card security measures and protections from Mastercard and Visa will apply to the AI payments. Additional guardrails include allowing users to set payment limits, specify permitted transactions and seek user approval.
AXS’ upcoming service removes the need to manually check for outstanding bills, log into multiple apps, and key in one-time passwords to make payments.
In a demonstration at the Singapore FinTech Festival at the Singapore Expo on Nov 18, Visa showed how a customer could type in a chat box to ask an AI agent on a mobile phone to pay a telco bill. The agent checks with the customer the amount and the card to make the payment. This triggers a biometric check to verify the customer’s identity using his face to complete the transaction.
AXS, Visa and Mastercard did not say which AI agents this capability will be available for. But AXS said that consumers could ask an AI assistant in their banking app or digital wallet to help with bill payments.
ChatGPT maker OpenAI said on Nov 13 that it was also working with American Express, Mastercard and Visa to let its dots AI agent make payments using the payment companies’ cards. This would initially be available only to US customers.
Abhijeet Ramesh, Visa Asia Pacific’s head of innovation and growth products, said that safeguards are baked into agentic billing payments using Visa credit cards and debit cards.
For example, only a legitimate agent verified to pay on the customer’s behalf can initiate a transaction. Explicit customer approval must also be given with biometric verification for the transaction to be completed.
The customer’s card details are also not shared with the agent for verifying and making transactions. Instead, a random code, called a token unique to the customer’s device, and a separate one-time code unique to each transaction, are shared.
Even if stolen by hackers, the token and one-time code cannot be used for any other payments, including to other billing organisations. Payment of a different amount from what was approved by the customer will also be blocked.
“We build these system-driven controls to make sure an agent can’t just do whatever it wants to and that a user is always in control,” said Ramesh.
Even so, as agentic payments are still in their infancy, Visa is working on how it can help banks, which eventually disburse funds for customers’ transactions, with more checks and balances for agentic payments, Ramesh added.
This includes updating banks on agentic transaction data so that banks can develop different rules to manage the risks of different types of payments to different types of organisations.
For example, based on specific trends in billing data, banks could set limits on how many transactions to a particular billing organisation can be allowed in a month.
Visa will use its existing framework for resolving disputes on card payments to handle disputes on payments made with AI agents. On top of that, there is more transparency with a tamper-proof record of each action an AI agent takes towards making a payment, which can be used for resolving payment disputes and determine if an AI agent is at fault for making an unintended transaction.
“We are also continuously looking at our dispute resolution framework and rules, which are quite expansive, to see what additional changes we need to make for agentic payments,” said Ramesh. “It is still early days, so we haven’t made any changes yet, but we are actively working on it.”
