Autonomous AI agents have been lauded as the future of online shopping, but one tech giant is pushing back even as payments giant Mastercard fetes its success completing what it calls Australia’s first “fully authorised and secured agentic transaction”.
That transaction – in which an AI agent used a Commonwealth Bank of Australia (CBA) debit card to buy movie tickets from Event Cinemas – was processed through payment technology company IPSI and completed using Matilda, the sovereign AI model from Melbourne firm Maincode.
The purchase was “fully authorised with cardholder consent,” Mastercard said, adding that it, the CBA, and Event Cinemas all got transaction updates thanks to Agent Pay – a global AI agent framework that is now expanding to the APAC and Australia regions.
A second transaction saw an AI agent asked to find and book accommodation in Thredbo, NSW – demonstrating how the technology can respond to user requests to complete a range of purchases on their behalf.
“Agentic commerce represents one of the most profound shifts in consumer behaviour we’ve seen in decades,” Mastercard Australasia division president Paul Monnington said in announcing AI-related investments including a Sydney agentic commerce team.
“Our role is to ensure this future is built on trust, security and transparency – the same principles Australians expect from card payments."
AI is learning to do the shopping for you
ChatGPT’s runaway popularity has primed Australians to tap AI technology for their shopping, with a PayPal survey last year finding that 48 per cent used AI assistants for online shopping searches and 61 per cent would trust AI’s product recommendations.
The volume of e-commerce traffic from generative AI (genAI) platforms increased 693.4 per cent during November and December compared with the same period in 2024, Adobe recently reported.
Significantly, mobile purchases made up 56.4 per cent of all online spend – the first time that figure had passed 50 per cent – but genAI queries were three times more likely to come from a desktop, suggesting AI-driven buyers were weighing options carefully.
Recognising an opportunity, AI giants have redoubled their efforts to help e-commerce giants to standardise product descriptions, transaction details and other information so that transactions can be fully automated across every step of the supply chain.
The goal is to get shoppers to commit to purchases faster by empowering them to describe what they want, then have AI agents identify multiple options and automatically buy the products – all without leaving the user’s AI chat session.
Agent Pay challenges the likes of Google’s Agent Payments Protocol (AP2), Microsoft’s recently announced Copilot Checkout, Visa’s Trusted Agent Protocol (TAP) and ChatGPT’s Instant Checkout – which it launched pre-Christmas with Etsy and Shopify.
Mastercard says it completed Australia's first AI agentic transaction. Photo: Supplied
Having autonomous AI agents making purchases on your behalf “is not theoretical”, open banking researcher Dylan Thiam recently observed, warning that while agentic commerce “seems like a risk management nightmare” it’s already too late to stop.
“As these systems become more integrated and more secure,” he wrote, “people will delegate even more.
“First for convenience, then for optimisation, and eventually because not doing so will just seem inefficient.”
But some are pushing back
Enthusiasm over the promise of agentic commerce has grown rapidly, with a recent analysis from consulting firm McKinsey calling the technology “a seismic shift [and a] rethinking of shopping itself in which the boundaries between platforms, services, and experiences give way.”
AI, McKinsey believes, “moves us toward a world in which AI anticipates consumer needs, navigates shopping options, negotiates deals, and executes transactions, all in alignment with human intent yet acting independently via multistep chains of actions.”
Its adoption could drive agentic commerce revenues to $1.4 trillion ($US1 trillion) in the US and $7.1 trillion ($US5 trillion) by 2030, McKinsey predicts – but to reach that point, proponents of the technology still have a bit of convincing to do.
Last August, perhaps wary of the problems created by rampant deployment of data-scraping bots – which by some measures account for more than half of all internet traffic – Amazon was spotted blocking AI firms from scraping data about its products.
Ironically, Amazon’s rival offering, called Buy for Me, recently drew fire from third-party brands upset that its own AI agents were scraping their data without permission.
Other firms have followed suit, with ImmuniWeb checking the policies of 1,807 major websites and finding that AI bots had become persona non grata – suggesting that many online properties are likely to be equally sceptical about giving AI agents free run.
More recently, eBay announced that from 20 February it will block users from using “buy-for-me agents, LLM-driven bots, or any end-to-end flow that attempts to place orders without human review” – explicitly calling out AI agents for the first time.
Analysts see the increasingly hostile reception for AI agents as a sign of growing protectionism from e-commerce brands, which all see the promise of agentic commerce but don’t want third-party rivals scraping data for their customers.
Whichever standards ultimately prevail, “agent-powered commerce could fundamentally change how our guests engage with their favourite experiences in the future,” said Andrew Turner, group general manager with Event Cinemas parent EVT.
“Being part of Australia’s first authenticated agent‑led transactions shows how AI could one day simplify planning a night at the movies or a trip to the snow, while keeping transparency and security at the core.”
