Latest / TeqTalk with Jas / You Won't Click 'Buy' Anymore; AI Is Doing It for You?
Transcript
- Teqfocus: Imagine it's Friday evening. Your daughter needs new running shoes for soccer practice on Monday. Instead of opening 10 browser tabs, you simply tell your AI assistant Find me the best running shoes under $120. Size 6, highly rated, arriving before Monday. Within seconds, it compares hundreds of products, checks inventory across retailers, verifies delivery times, and completes the payment. You never opened Google, you never visited Amazon, you never clicked a single ad. The AI did everything on your behalf. That might sound like the future, but it's already beginning. In July 2025, traffic from AI chatbots to US retail sites grew 4,700% in a single year. That's not a stray statistics. It's an early signal commerce is being fundamentally restructured. For 30 years, online shopping followed one pattern human search, human compared, human decided, and businesses built entire industries around convincing people to click. Agent e-commerce changes the fundamental assumption. The person may no longer be making the buying decision. Increasingly, software will. And once software becomes the buyer, new questions will appear. How does an AI know which merchant to trust? How does it verify a seller is genuine? These questions matter more than which AI model is smartest. Because the future of commerce isn't about intelligence, it's about architecture. The world isn't converging on one answer. Distinct architectural philosophies are emerging and each one believes it's found the right way to build this. Think about smartphones. Apple didn't win by copying Android, and Android didn't win by copying Apple. Both succeeded by solving the same problem differently. Something similar is happening in commerce right now. The open ecosystem, the integrated ecosystem, and the open network. Let's start with the open ecosystem. Imagine you are Google. Millions of people use your search engine, but you do not own most of the commerce journey. Not the retailers, not the payments, not the warehouses. So an AI shopping agent has to discover a product through one company, verify inventory through another, and complete the payment through third. The challenge isn't intelligence, it's coordination. That's why Google, Anthropic, OpenAI, Stripe and Shopify are each building protocols. Anthropics NCP, Google's A2A and Universal Commerce Protocol, OpenAI and Stripe's Agenti Commerce Protocol. So agents, merchants, and payment providers can talk to each other. No single company owns everything. Walmart is the proof. Its Sparky system runs on NCP, letting customers Employees and supplier agents share one data plane without owning the whole stack the way Amazon does. But that's not the only American one of the strongest examples of a completely different philosophy also comes from the US. Amazon. Amazon already owns most of the journey. Discovery, payment, delivery, support without leaving its own ecosystem. So Amazon isn't asking how do independent companies work together. It's asking if we Already own the journey. Why not make AI native inside it? And that's the integrated ecosystem. The platform becomes the bridge itself, which is exactly why Amazon blocked outside AI crawlers rather than opening up to them. And this is where China gets interesting. China isn't copying the US, it's much closer to the Amazon's philosophy than Google's. Alibaba. Tencent and Byte Dance already run ecosystems that combine commerce payments and messaging in one. Alibaba's Quen Assistant is now integrated across Taobao and T Mall's 4 billion product catalog, past 300 million monthly users. Their goal isn't just a smarter chatbot, it's making the entire journey invisible because AI is woven into every interaction they already control. One believes Collaboration across independent companies wins. The other believes owning more of the journey wins. Neither is wrong. They are different architectural bets, third model. And it may surprise you the most. It isn't trying to become another Google, Amazon, or Alibaba. And that brings us to India. India looked at the same problem and asked a bigger question. Not who should own the journey, but what if nobody owned it? Think about how email works. You can send messages from Gmail to Outlook because both follow the same shared standard with no single company in the middle. India is building commerce the same way through the Open Network for Digital Commerce, ONDC. It isn't another marketplace, this is a network underneath commerce. Already past 200 million transactions with any retailer, logistic provider or payment app free to plug in. It's the same philosophy that built UPI now aimed at AI. India is going one step further. At the India AI Summit in Feb 2026, industry leaders called the same playbook Apply to Agents, an Aadhar for Agent Identity, a UPI for agent payments, an ONDC for interoperable commerce. So every agent transaction carries a cryptographic auditable identity from day one. On that foundation, Flipkart became a founding partner of Google's own universal commerce protocol. And Big Basket went live on Chat GPT through RazorPay and UPI. So zoom out. This isn't a competition between the countries, it's a competition between the architectures. Open System says collaboration wins, integrated ecosystem says ownership wins, open network says infrastructure wins. For 30 years, business optimized for SEO, helping humans find a website. But agents do not browse the way humans do, they do not scroll. They do not get swayed by branding. They consume structured data and recommend one answer. That's why a new discipline has a name now. Generative engine optimization, GEO. And it's volatile. With the sources and AI sites shifting 40 to 60% month to month, far more than Google ranking ever moved. Imagine two companies selling the same running shoes. One has beautiful branding, the another one has complete specs and verified delivery dates. A human might prefer the first. An agent trusts the second. It doesn't rank incomplete data lower, it just never recommends it. India already has a number of what that costs. Fragmented product data is estimated to lose Indian retailers roughly 5,000 crore a year in discovering alone. But discovering is only half the story. Once an agent decides what to buy, it still has to answer. Can this transaction be trusted? Each architecture answers differently. The open ecosystem leans on contracts. The EU's product liability directive treats AI as product under strict liability starting this December. Integrated ecosystem leans on platform control. China requires algorithm registration before AI commerce service can launch. Different philosophies, same underlying question. And trust isn't just identity, it's money. Card networks were built for humans swiping a card, not software completing thousands of microtransactions a second. And every architecture rebuilding payment rails to match. Visa's trusted agent protocol issues a cryptographic mandate, proof of the human authorized the purchase. Google's agent payment protocol already has more than 60 partners, including MasterCard, Visa, and PayPal. But the open networks are furthest ahead in practice. Alipay already moved 120 million agents' payment in one week, and India routes agent purchases straight through UPI, the same rail that clears billions of human transactions a month. And then there is security. Most assume AI reduces fraud. Attackers use it too. Palo Alto Network has already documented indirect prompt injection attacks against Google's own universal commerce protocol, where a malicious page hijacks an agent's Cart mid-transaction, it looks like a clean, successful order. So, cybersecurity used to protect people from malicious the next decade may require protecting AI from malicious AI. So the real question isn't which AI model should we adopt. It's actually whether an agent evaluating your business has enough trusted information to confidently recommend you. Let's go back to where we started. A father asked an AI to buy his daughter running shoes. It just works. But trust has to be built first and again and again at every stage. Agent recognization. Is this really an AI acting for a real person? Discovery. Is the product data actually true? Evaluation. Can the agent trust it enough to choose with no human checking? Payment. Does the money move safely? With proof someone real said yes? Fulfillment. Does the order actually ship the way it promised? Accountability. When something breaks, does anyone actually own it? Six checkpoints, one requirement running through every single one. That's the actual product being built right now. Not a smarter model, not a slicker checkout. Trust. Engineered well enough that nobody has to think about it. And trust like that doesn't scale slowly once that's real. UPI went from a handful of transactions to billions a month in under 10 years. The moment people trusted it enough, Stop questioning it. So whether you are an executive setting the strategy, a senior leader defending the budget, the product leader building it, a seller trying to get discovered inside it, or a buyer deciding what to trust with your money, the question isn't whether agent e-commerce scales, it is how soon and whether the phase you own it ready to be trusted at that speed. If you like the episode, please subscribe to Tech Talk with Chess.