One of the biggest differences between being rich and poor is that people have to work your Things. When a person leads to the middle class, he can get a cleaner or take his clothes to a dry cleaner. When they get rich, they can find the driver, a private chef – a whole delegation is dedicated to facilitating their lives.
There is a growing excitement about the rise of AI -powered agents that work by consumers – not only during the discovery of the product, but also for purchase. Chat -looking questions on GPT can only represent 1 % of the people on Google, but 1 % has a lot of searching in the Great Global Search Market, which potentially has millions of revenue.
Originally considered as digital documents to simplify the search, these agents are now making original purchases, and they are doing this without ever being handed over to the buyer.
You can like
Large payment players are the basis for AI -based trade. The visa has recently launched a digital credentials hub to find new identification models for agent -based transactions, while Strip has confirmed that it is developing safe transactions for AI agents. And just yesterday, Google revealed AI Agent’s checkout plans for purchase – a move that confirms this shift is no longer speculated, but is an ascension.
But there is a complex picture beyond the hope. What happens when AI agent makes purchases for you? And more importantly, what can be wrong?
Chris Jones
Social Links navigation
Managing Director, PSE Consulting.
Really what is happening behind the curtain (usually)
Let’s clearly: There is no “quality” for agents purchasing – this process is still getting ready, and different platforms take different approaches. He said, here’s a common flow that we have witnessed in the initial implementation.
When a user uses an AI agent for purchase, this process is surface, but technically complicated. First, the user saves their payment card details, including full PAN, CVV, expiration, billing, and delivery addresses with their selected AI platform.
Buyers are unlikely to buy very cheap items like pizza or expensive items like a new car. They will probably not use it for goods with heavy visual force, where a part of it is browsing unless something gets hit with you – clothing is not the best example. Possibly, they first use agent AI to help them decide between relatively expensive products that are difficult for non -experts to understand: Let’s use a good pair of Bluetooth headphones for example.
The agent, which can be strengthened by the AIA measures of the Chattagot, Google, Tacotok Shop, or Amazon, uses the natural language to respond to the buyer’s request. Just like the shop clerk, it will ask questions to improve the results: How much do you want to spend? Do you want headphones in more ear or ear? Are there any features such as noise cancellation or waterproof you need? It can then improve the results and offer purchase options.
Once the shopper decides, the payment process begins that will mostly be hidden for the buyer:
- The buyer lives in the AI interface and never visits the merchant’s site.
- The “Buy” command agent inside the UI enjoys the checkout form on the merchant’s site to autofell.
- The merchant gets full details of the card as if someone is typing a human shopper.
- The agent offers the order, and is sent by both the confirmation agent and the merchant.
Critically, the trader is unlikely to know that he is dealing with an agent rather than a human being. This introduces the risks, because if something goes wrong – a wrong item, delivery mixup, or pricing error – the buyer must have to solve it directly with the merchant, even though he has never interacted with the merchant website.
In other words: Don’t talk to me – talk to my agent.
Famous dangers (so far)
There are already many emerging disadvantages.
Security risks: In January 2025, the Chinese AI platform Dippec was hacked, which exposed the consumer stored credentials. Centralizing payment data in AI agents makes them profitable targets.
The sensitivity of scams: Delers can design the SITE sites of tricks, especially agents, to complete the fake checkout.
Irritation in Liability: If an agent misrepresenting an order or entering the wrong details, it is unclear whether the AI provider or the user imposes responsibility.
Poor compatibility:
- It cannot support alternative payment types such as PayPal, digital wallet or bank transfer (which is about 45 % of the EU commerce volume in the European Union).
- It cannot easily handle additional checkouts (such as selection, delivery slot).
- Card reduction struggles, especially in international transactions, where the reduction rate can be 5 to 30 % anywhere.
In markets like the European Union or Japan, the legal requirements around a strong customer verification (SCA) mean that consumers must approve each card transaction, which causes AI -led flow trouble or non -compliance.
Big Photo: Are we witnessing a commerce revolution?
Beyond immediate risks and supplies, the rise of agents raises basic questions about the digital trade structure.
Will this model get traction with users? It can shine like sound trade and Amazon’s dash buttons, which failed to end due to confidence and use. Or it can burst, such as the rise of markets or the app to buy mobile. The answer depends on how much users are at the convenience, and how well AI agents can overcome confidence and control issues.
If AI agents become a priority interface for e -commerce, the web as we know that this piece can be. Why go to a merchant site when your agent can work? This change can lead to a model context protocol (MCPS). Some merchants can respond by blocking the leading agent IPS or designing checkoutflows that frustrate automated systems to force direct interactions. Industries like marketing will be mainly changed because engaging with AI agents is more important than humans.
Meanwhile, platforms like Chat GPT will need to find ways to earn its new influence. This may mean that traders charging referral fees, gives rise to the appearance of a new SEO for-AI environmental system. But new questions are introduced about confidence in such money: If your agent is taking commission from traders, how neutral is its recommendations?
When we stand on the edge of this change, one thing is sure: the infrastructure of digital payments is being re -written. Now the question is, are consumers – and merchants – ready to follow their agents in this new era?
And I, for one, will be looking closely – either directly, or through my agent.
Link!
This article was developed as part of Tech Radarpro’s expert insights channel where we present the best and bright minds of the technology industry today. The views expressed here are of the author and it is not necessary that they belong to the Tekradarpro or the future PLC. If you are interested in getting more information here: https://www.techradar.com/news/submit-story-tory-techradar-pro
