Ben Thompson on Agents, Amazon’s Logistics Moat and Who Owns the Customer When the Agent Buys, on TBPN
Written by an Ethmos research agent · shared with you
Thompson says Amazon's logistics moat, not Shopify's platform model, will decide who wins agent-mediated commerce.
- Amazon's logistics moat wins: Thompson says agent-mediated shopping won't dislodge Amazon because ChatGPT can't even link to it, yet delivery certainty and easy returns keep him buying there anyway.
- Shopify can't close the loop: Shopify stays asset-light (its stock cratered when it tried logistics) and works across every platform, but by design can't solve the physical fulfillment problem that keeps consumers on Amazon.
- Walmart, not Shopify, is the real rival: Thompson calls Walmart's app and one-hour in-store delivery underrated; OpenAI's checkout integration favored Walmart but reportedly saw smaller carts and abandonment versus Walmart's own site.
- Reframe agents as pain relief, not productivity: Thompson argues people don't want to be more productive — pitch personal AI as eliminating annoying chores (returns, mail, subscriptions), not aspirational efficiency.
- People love shopping, don't want it automated: he pushes back on flight-booking and shopping-agent demos, saying consumers enjoy the research and even the complaining and resist full automation.
- Auto-cancel agents already causing support headaches: Stratechery sees new customer-support tickets from agents auto-unsubscribing or canceling recurring subscriptions via inbox scans, a real-world friction point for subscription businesses.
Deep dive
Personal assistants as pain relief, not productivity
Thompson, drawing on a decade of managing a human executive assistant for Stratechery, says most people don't know how to use a personal assistant and it took him years to learn. His central framing for AI agents like Meta's Muse is that they should be marketed as eliminating annoying tasks — unopened packages, unprocessed returns, unpaid parking tickets — rather than as productivity tools, since he says people simply don't prioritize being more productive in their free time; employers care about that, but off the clock people just want to relax. He predicts agent adoption will spread the way social-video content spreads — through people showing what the tools can actually do — rather than through aspirational pitches.
Building "Geckobot" and the home-inventory app
Thompson describes building an internal AI assistant he named Geckobot, after a metallic gecko piece on his wall, to take his task requests, photos and reminders; separately, a home-inventory system grew out of a cross-country move from Taiwan that left him unable to find anything his own staff had put away. The more striking result, he says, was watching his human assistant's morale improve: the bot let Thompson log weekend tasks with photos without guilt, and the assistant could batch and execute them Monday morning, calling it one of the best weeks of work he'd had. Thompson uses this to argue AI's near-term value is filling countless tiny "software-shaped holes" in workflows — small, disposable internal tools that don't need to be shipped as products — a capability he says even technical people underestimate.
Toward humanoid and niche robots
Both hosts and Thompson converge on physical-world robotics as the more exciting frontier than software agents. Host Jordi Hays describes hiring a human driver after bad fires near his home wrecked his commute, and realizing the value wasn't the driving itself but everything else — parking, errands, dry cleaning — concluding self-driving alone (even at Tesla's reported 90%+ autonomy) solves the driving task but not the broader job he needed done. He envisions an end-state with a general-purpose humanoid maintaining a fleet of niche single-purpose robots (window cleaners, pool cleaners, a mail-processing robot that slices, photographs, and routes mail), while Thompson says paying a person to handle those chores is simply living a decade early.
Shopify versus Amazon in agent-mediated commerce
Pressed on whether Shopify merchants could fill Amazon's long-tail inventory gap, Thompson says the dispute comes down to Amazon's logistics moat: reliable delivery and easy returns. He notes AI chat tools structurally can't link directly to Amazon, forcing users back to Amazon's own ad-heavy site anyway, which he still prefers because he trusts its fulfillment. Shopify, having previously tried and abandoned a logistics push (visible in its stock chart), instead stays asset-light and cross-platform — a smart business model, in his view, but one that can't address the fulfillment reason people default to Amazon, leading him to conclude Amazon will effectively win this dispute.
Walmart as the overlooked rival
Thompson calls Walmart's under-the-radar strength in e-commerce a blind spot for tech audiences, citing its in-store navigation app and one-hour delivery from physical stores. He says OpenAI's checkout partnership favored Walmart initially, but that Walmart's own site converts better (larger carts, less abandonment) than the ChatGPT-routed experience, reinforcing that owning the full customer experience — as Amazon and Walmart do — matters more than agent integrations.
Consumer psychology and subscription-agent friction
Thompson pushes back hard on demos of AI booking flights or shopping automatically, arguing people enjoy the research and even the complaining involved in shopping and travel decisions. On the money-saving/subscription-cancellation use case, he notes real operational harm: Stratechery has seen a wave of support tickets from users whose agents auto-clicked email unsubscribe links or canceled recurring subscriptions without users' intent, since unsubscribing from email and canceling a subscription are legally and technically distinct actions.
Briefs like this, for everything you follow.
Ethmos puts research agents on your coverage universe — every debate that touches your names, every executive appearance you'd have missed — and files briefs like this one every morning.
Start freeThis page is an AI-generated summary and analysis prepared with Ethmos and shared by an Ethmos user. It does not reproduce the original programming. The underlying episode and its recording remain the property of their respective creators, and all show and company names are the property of their respective owners.
AI summary. May contain errors. Not investment advice.