Meta’s Muse has been around for a month now, and its rise in that time has been meteoric, hitting five million downloads in its first 22 days. Muse is built to understand the goals of each of its users and to perform tasks on their behalf – and that includes paying for items. It has integrated Stripe’s Link payment system, including one-time-use card functionality for eligible transactions, with Shop Pay support also on the cards.
Not to be outdone, OpenAI has announced the launch of its rival Dots agent, which performs similar functionality – and there are signs that it is preparing a payment product called ChatGPT wallet, which would let users store payment cards for use while carrying out tasks.
Haven’t we seen this already?
OpenAI has already experimented with functionality enabling users to buy directly from merchants in ChatGPT; Instant Checkout lasted just a few months before the company quietly dropped it in March, stating that it did not “offer the level of flexibility that we aspire to provide”. The key difference with the latest approach to AI shopping is that these agents do not seek to replace retailers’ checkout journeys. Instead, they act as intermediaries, carrying out purchases on users’ behalf while transactions are still completed through merchant payment rails.
So, does this new approach mean that AI shopping is finally ready to scale?
Chris Jobes, Managing Director at expert payments consultancy PSE Consulting, remains unconvinced, warning that agent providers becoming “online wallets” that sit between consumers and merchants with whom they have no established relationship increases risk for both merchant and customer.
“Giving agents access to consumers’ payment credentials makes buying smoother, but it also concentrates risk around identity, authority and fraud,” he said. “Merchants have less visibility of who is behind the purchase and when goods fail to arrive or a payment is disputed, more of that risk shifts to the wallet layer. An agent that can pay is also a new target for fraud. If criminals can impersonate an agent, manipulate its instructions or exploit weaknesses in how it proves its authority, they could trigger purchases the customer never properly approved.
Advertising muddies the waters
He also highlighted the fact that advertising may muddy the waters. If agents are funded by commerce and advertising, the consumers are unlikely to trust them to make impartial purchasing decisions on their behalf. This, Jobes argues, means that transparency and accountability are non-negotiable – and he doesn’t see that trust piece happening until agentic commerce is more regulated.
“To deliver a frictionless consumer experience at scale, the industry needs common standards for identifying agents, verifying their authority, authenticating payments and resolving disputes, with clarity over who carries the risk when an agent gets it wrong,” he said. “Until then, getting an agent through checkout is a short-term technical fix rather than proof that agentic commerce is ready to scale.
“It is hard to see this as a model that will scale, but for now it is the only credible route to giving consumers the feel of agentic commerce,” he added.
In other words, AI shopping has arrived before the industry has agreed the rules. The agents may be ready, but – as Jobes makes clear – the plumbing behind them is still very much a work in progress. For now, the technology may have solved the checkout journey, but not the commercial framework needed to support it at scale.
Stay informed
Our editor carefully curates two newsletters a week filled with up-to-date news, analysis and research. Click here to subscribe to the FREE newsletter sent straight to your inbox. Why not follow us on LinkedIn to receive the latest updates on our research and analysis?
