Your next purchase might not be made by you at all — it could be made by your AI agent, negotiating, choosing, and paying on its own. Coinbase, Google, and Stripe have all launched protocols to make this real in 2026. Joining me to unpack it is Akshay Chopra, Payments Committee Co-Chair at the MENA Fintech Association and Managing Partner at 237 Ventures. Akshay, welcome to Wall Street to Mena.
Thank you for having me back.
Agent commerce, agent payments — everyone is throwing this term around. What does it mean in simple terms?
It is an evolution of AI. Right now, AI helps you make decisions. Agentic AI can actually make decisions for you and act on those decisions. If you are going to Vietnam, AI today can help you think about which hotel to choose. An agentic AI will choose that hotel for you based on your preferences and budget, make the booking on its own, and just send you the confirmation.
So it will do everything for you?
Yes — based on the rules you have provided.
How is that actually different from the digital payments we already use every day?
The payments industry has been trying to minimise the time someone spends in the payment moment — from cash to card swipe to tap to one-click. We have already reached the point where payments are invisible. If you book an Uber, you do not make a payment — it just happens automatically in the background. But the next level is where even the moment of commerce becomes invisible. You tell your AI you want to go home. It considers your budget, how urgent it is, your preferences — Uber or Careem, or maybe the metro because of surge pricing — and simply tells you: here is the car waiting outside, or here is your metro ticket.
Isn't it also scary to put all your information out there?
Absolutely. Data ownership and data privacy are things you have to be thinking about. Human beings start generating data from the time they are in the womb — your ultrasound images and vitals. Literally before you are born. You have to be very aware of the permissions you are giving different companies and different AI models on how they can use your data. I strongly recommend taking baby steps — check your phone's privacy settings, see which apps have access to which data, and if you use a chatbot like ChatGPT or Claude, check what data it is storing about you and refresh it regularly.
Salesforce just put shopping agents live inside chatbots. Are merchants even plugged in yet?
Merchants want to sell regardless of the channel. A lot of advanced merchants are already looking at LLM optimisation — the equivalent of search engine optimisation, but for AI models. How can an AI agent find my product, know it is better, and choose it for the customer? You are entering an era where machines are doing the shopping. The question becomes: how do you convince the machine to shop your product?
Where does this hit first — consumers or businesses?
I actually think the business sector has tremendous applications. Businesses already have procurement rules, supplier rules, budgets — very hard-coded process logic that AI loves. If you are a hotel and need to order 50,000 towels, you can tell your AI agent what you want. It knows your budget, your supplier criteria, their ESG ratings, their background checks. It can negotiate with those suppliers' agents, find the best deal, raise the purchase order — everything done.
Doesn't that also open the door to AI fraud?
Absolutely. In this world, fraud evolves from someone impersonating you to someone manipulating your AI — pretending to be your agent, or fooling your AI into thinking you want to purchase something, or increasing the spend limits you set. In this new era of fraud, you have to verify whether it is genuinely the individual or an authorised AI operating with the right intent and the right controls.
So AI agents should have their own ID checks?
There is a lot of great work being done by Visa, Mastercard, and Stripe on standards for this. If you authorise your AI agent, what does that look like? How is it transmitted to the merchant? Is this the right agent? Does it have authorisation to pay at this level? Is the merchant legitimate, or is it trying to manipulate the AI through false reviews? Those are the questions the standards are being built to answer.
78% of banks expect fraud to spike from AI agents. What rules need to exist to protect consumers and businesses?
Let me give you a real example. My mother's birthday is coming up. Right now I go to Amazon or walk around a mall, think about what she likes, and buy her a gift. In the future, an AI agent could look at her Instagram, her interests, her Amazon history — with permission — find the best gift within my budget and deliver it. But what if the gift is wildly inappropriate? Or it spends twice as much as I intended? Or it breaks conditions I had set? Whose fault is it — mine for authorising the agent, the AI company's, the bank's, or the card network's? When the human is completely removed from the transaction and it is machine to machine, the entire question of liability changes.
Who is actually liable when an AI agent makes a mistake?
Visa and Mastercard have been setting clear global rules for liability since the 1950s — when is the merchant liable, when is the cardholder, when is the bank. I fundamentally believe they have the capability to take that same skill set into this new world. There is no easy answer. It will be a complicated set of rules and liability shifts — but it will be built, because it has to be.
Thank you so much. It is very interesting to hear all of this today. Thank you for being here with us.
Thank you.
