AI shopping agents are officially becoming a thing. Big payment providers are building the infrastructure to let digital agents handle your purchases, compare prices, and even snag the best deals, without you lifting a finger. While it all sounds extremely exciting, here’s a bummer – most people aren’t exactly lining up to hand over the shopping reins to AI bots.
AI agents doing your shopping, while other AI tools participate in payment processing on the other end? When I first heard about the concept of AI-to-AI economy a couple of years ago, it seemed like a matter of the distant future. Yet, it’s not sci-fi anymore; payment and fintech companies are already building real infrastructure for it. But based on trust data, lots of people are still hesitant to hand over control. Here’s what’s going on, why people are cautious, and how payments firms are racing to make it safe and seamless.
Why People Are Skeptical About AI-Powered Shopping
According to Worldpay’s 2025 Agentic Commerce Report, only around 40% of consumers say they would let an AI shop for them, while another 33% are “open to the idea.” That leaves a big chunk of the population hesitant or flat-out skeptical.
So, what’s holding people back? For sure, some share of people just love the thrill of shopping and won’t be eager to give that joy away to any other person, not to mention machines. However, most of us would gladly simplify at least part of our routine shopping duties. And yet, delegating it all to AI is not that attractive to most people (yet?). It turns out trust, or the lack of it, is a major hesitation factor. Half of the respondents cited fears about fraud, identity theft, or unauthorized purchases:
-
55% worry about identity theft;
-
55% fear incorrect or unauthorized purchases;
-
53% are concerned about fraud;
-
51% feel they could lose control over their finances.
And demographics play a role, too. People under 35 are more willing to trust AI with purchases, roughly 50% would hand over their shopping list to AI little helpers, while just 15% of those over 55 feel the same.
So, what would make consumers more comfortable? Worldpay’s survey points to a mix of security and control:
-
54% want strong fraud protection;
-
50% want the ability to cancel a purchase within 24 hours;
-
49% want a review step before the AI finalizes a purchase;
-
43% feel that having human customer support is important.
Only 5% of people globally say they have no concerns at all about agentic commerce, which shows that while the technology is promising, widespread adoption isn’t guaranteed without clear trust measures.
Meanwhile, Payment & E‑Commerce Providers Are Betting Big on Agentic Commerce
Even though many people are wary, payment providers aren’t waiting around. Payment providers are racing to build the systems, identity verification, and “digital agent passports” needed to make AI shopping safe and transparent.
Big Networks Building “Agent‑Ready” Payment Systems
-
Mastercard launched its Agent Pay solution, which is designed to let verified AI agents make payments on behalf of users, in the US and UAE with global rollout in plans. It uses agentic tokens — cryptographically secure payment credentials tied to a specific, trusted agent.
-
Visa introduced Visa Intelligent Commerce, a set of tools (APIs, SDKs) that let AI agents act within limits set by users. These tools support spending controls, fraud protection, and secure identity attestation.
-
PayOS + Mastercard: PayOS (a payments and agentic services platform) has completed a live transaction using a Mastercard Agentic Token. This isn’t just theory anymore, agentic payments are processing real money on existing rails, though with constraints on what an agent can buy and when.
-
Ant Group (Antom): They’re piloting agentic AI payments both with cards (via tokenised card rails) and digital wallets. Their system includes a mandate model: agents carry a “permission slip” that links user intent to the transaction, which helps with security and dispute resolution.
Platform + Protocol Partnerships Propel Agentic Payments Forward
-
Stripe & OpenAI jointly developed an Agentic Commerce Protocol (ACP), which is being used for “Instant Checkout” inside ChatGPT. That means if you ask ChatGPT to find something, it can actually send a purchase to checkout without exposing your full payment credentials.
-
PayPal & ChatGPT: PayPal adopted the Agentic Commerce Protocol, so ChatGPT users can check out through PayPal’s network (bank accounts, credit cards, balances) using familiar protections (like dispute resolution and buyer protection).
-
Balance (a B2B fintech infrastructure company) launched a Model Context Protocol (MCP) Server, letting AI agents talk directly to Balance’s payments, receivables, and credit APIs. This is super relevant for businesses that want to automate financial operations using conversational AI.
Regional and Emerging Infrastructure Moves
-
India is experimenting with agentic payments via UPI (Unified Payments Interface). A pilot with NPCI (India’s payments body), Razorpay, and OpenAI lets ChatGPT users in India pay through UPI inside the familiar chat experience. Agents can complete purchases using pre-authorized UPI “Reserve Pay,” meaning users still retain control while agents act autonomously.
-
Kite AI, backed by PayPal Ventures and others, is building a blockchain-native infrastructure for agent identity and payments. Their tech includes “Agent Passport” for identity, programmable payments, and native settlement for a future where agents don’t just order things, they handle microtransactions and on‑chain commerce.
Where Does This Mismatch Leave Us?
Payment players are gearing up for a major shift. The biggest names in the industry are investing heavily in the rails that would let AI agents operate safely. This includes everything from identity layers to specialized payment tokens designed for autonomous transactions, and costs millions.
But consumer trust still hasn’t caught up. Even with the underlying tech maturing fast, many people remain uneasy. Worries about fraud, unclear decision-making, and losing visibility into what an AI is doing with their money are still very real.
Clear rules will make or break adoption. For agentic commerce to work at scale, it’s not just about smooth payments, but also largely about knowing who’s accountable if something goes sideways. That’s why new systems focus on linking user intent to every transaction and formally registering the agents involved.
Will there be any penalties for e-commerce or payment providers who let their AI agents mess with people’s money? Who knows? Frankly speaking, regulators are also not so quick to react to the quickly evolving payment and shopping landscape. Given the legal grounds for AI shopping are clear and transparent, far more people are likely to test the new waters.
And yes, this opens the door to new business models. From AI-powered shopping companions and smarter social commerce to automated B2B purchasing, the shift could reshape how both consumers and companies buy things.
Bottom Line
Payment giants are rolling out the plumbing that lets digital agents check prices, compare products, and even complete purchases on your behalf. Despite all the innovation happening behind the scenes, people hesitate. It’s not too shocking, as their valid concerns are rooted in trust, or rather absence of it.
The experience might be promising and refreshing, but consumers still want reassurance that AI agents won’t misuse their data or make decisions they can’t undo. So, if agentic commerce is going to go mainstream, it’s not enough to have powerful AI agents: we’ll need transparent protocols, strong identity verification, fraud protection, and ways for people to stay in the loop. Until that, most shoppers are just not buying it.
