Week 13 - AI rewires ecommerce playbooks: Shopify’s agentic bet, Amazon’s tech war, and the coming checkout shakeup

Max: Welcome to the Checkout Point, your quick dive into this week's e-commerce buzz with me, Alex, powered by Blicket. These 20 minutes pack the latest trends, news and insights. Let's navigate the digital marketplace together. Ready? Let's go! Rachel: Hey, hey, welcome back to the Checkout Point. Speaker 3: Good to be here, Max. If you're trying to run e-commerce at scale and actually keep margins in 2026, this one's for you. Rachel: Yeah, this is the Stop Buying Shiny AI Tools You Don't Need episode. I spent years in CRO watching this exact pattern, teams burning budget on features that never move the dial. Speaker 3: Please, my procurement PTSD from Walmart is already flashing. I've seen six-figure AI contracts for problems you can solve with a spreadsheet. Rachel: So first up, we're hitting AI and e-commerce, how Shopify's Google AI, tools like Alia, and agentic pop-ups look great in demos, but we're going to map them to real 90-day KPIs. Pick one wedge, not 10 toys. I've seen this movie before, and it doesn't end well. We'll close today with logistics and inventory-how AI agents quietly move into your back office and one small inventory experiment you can try next week that doesn't require dropping seven figures-AI Speaker 3: But first, let's start where all the hype decks start, and almost none of the dashboards do. Rachel: in e-commerce and actual strategy, not the hype version. Speaker 3: Let's get into it. Rachel: Roll segment one. Three easy ways to get your questions to us: Check the description for our web form, text us at seven four seven two nine three four six one two or call the same number to leave a voice question. However you reach out, we can't wait to hear from you. Picture this: we're at Berlin Expo two thousand twenty six. You walk past a store, but it's not really a store; it's just AI doing what I spent literal years trying to hack with split tests in AB frameworks. Speaker 3: It's like a screen wall and some voice pods, right? Rachel: Exactly. No static homepage (and trust me, I've burned through conversion data showing how bad those are), no twenty link nav nightmare-you say, "I need a week's worth of outfits for meetings in the rain. And an AI agent just builds your cart across three brands. Speaker 3: Wow. And meanwhile, I'm watching every legacy retailer debate nav architecture while the rules completely shift. I mean, so Rachel: That's the gap right there. Expo made it super obvious. AI is eating the storefront and honestly, doing it better than most merchants ever could. Speaker 3: what's actually real for merchants rolling it out, not just Berlin Expo theater and week out timelines? Rachel: two buckets: Shopify's AI ecosystem, especially that Google AI integration reportedly driving 15 times more orders; two, tools like Alia's AI Pop-Ups and Pattern's discovery engine. That's where the action is right now. Speaker 3: Let's hit Shopify first. Fifteen X is huge, but what's the actual baseline we're measuring from, because I've spent my entire CRO career hunting these claims, and the devil's always in the funnel. Rachel: It's using Google AI as your front door. Instead of hunting through your site, someone types best running shoes for flat feet under $150 into Google. The AI pulls your structured product data from Shopify and ranks your best matches. Speaker 3: So it's like product listing ads, but now Google's AI is doing your merchandisers' job. If your data is clean, you win. If it's trash, you're invisible even with the model helping. Rachel: Right. And here's my philosophy from CRO days. Feed the algorithm your best stuff. Clean titles, real attributes, live availability. When you nail that product feed, your odds of being that top conversational recommendation go way up. Speaker 3: Mm-hmm. Here's where I get annoying. Fifteen acts from a ten order baseline to a hundred and fifty is mathematically huge but operationally still small; I need the full funnel, Blended CAC, repeat rates, whether it's actually incremental or just shuffling existing demand across channels. Rachel: I'd call that proof of direction, not a forecast. But here's the operator check I always run: is it incremental or just cannibalizing your branded search? Because that's the question nobody wants to answer. Speaker 3: Right: Blended CAC, new versus repeat, margin after platform fees, because fifteen x looks great until you realize you're paying for it in rev share and your unit economics are worse. Rachel: Exactly; if your total orders are flat, but AI-driven orders are up fifteen x, you didn't win, you just shuffled the game around. Speaker 3: Right, second bucket. Agentic funnels on site. Twenty seven hundred Shopify stores running AI pop-ups, Dotdigital buying Alia. But I've seen too many smart pop-ups that are just dark patterns with UI makeup. Where's the actual difference in unit economics? Rachel: The dark pattern casino wheel, man. I've seen so much of that back in my CRO days it's wild how it still shows up with a new coat of paint. Speaker 3: Ha! ha! ha! I absolutely have. Stores use AI as an excuse to get aggressive about discounts and data grabs. Rachel: Speaker one: The best implementations use behavior and catalog data smart. If you've hovered on premium products, it offers a bundle builder, not a margin killing coupon. That's the difference between AI and just AI. Speaker 3: And Patterns forty per cent. revenue jump, same question-baseline metrics, new versus repeat, incremental or just optimization of existing traffic. Rachel: My guess is way better session level matching; dynamic collections; shop the routine flows; multi step quizzes powered by an actual model instead of hard coded if then trees that drive me nuts. Speaker 3: Instead of a static skincare category, it walks you through skin type, budget, concerns, then builds a cart. That's solid CRO, just now it's got a smarter brain doing it. Rachel: Exactly. We used to build these with long forms and decision trees. Trust me, it was painful. Now an LLM asks fewer, smarter questions and lands you on the right 3 to 5 SKUs. Speaker 3: Right, but here's what kills merchants. They want to test search, recommendations, Pop-Ups and Agentic all at once. You can't do that without torching your KPI clarity. I learned that the hard way. Rachel: No, Please don't. You'll turn your analytics into a dumpster fire. Max: I've cleaned up enough of those to know it's not worth it. Rachel: Exactly. I've seen that nightmare before. Max: Start where friction is highest and intent is clearest. For most brands, that's on-site search or post-click discovery. Both have tons of conversion juice just sitting there waiting to be picked. Rachel: So search or recommendations first, not Pop-Ups; Fix core discovery before you layer in the funneling tricks. Max: Yeah, Pop-Ups are Layer Two If someone searches black dress under a hundred dollars and your product data is a mess fixing search with AI is huge do that first Rachel: Let's ground this in real metrics: first ninety days, concrete, no vanity numbers. What's actually moving the P&L? Max: For AI Search, search to add the cart rate, search exit rate, and revenue per search session. If those don't move, you're just buying lipstick on a pig. Trust me, I've seen that movie before. Rachel: For recommendations? Max: Click through on Rec's AOV and attach rate on key hero products. Segment new versus returning. New shoppers should benefit most. Rachel: And for AI Pop-Ups or Agents? Max: Email capture quality. Actual down funnel revenue per subscriber, not vanity opt in rates. For Agents, percentage of sessions that engage with the Agent and convert. Skip the ego metrics. Rachel: I'd push further-ticket deflection if the agent handles support, and any spike in wrong item returns if matching logic is flawed. AI helped that tanks your RMA rate is actually a loss. Max: That's a good one. If your AI agent helped and your RMA queue explodes, that's not a win. Rachel: So here's the takeaway: AI is eating the storefront, but the smart play is one wedge, one KPI, 90-day window, no exceptions. If it doesn't move the needle, you kill it and move on. Max: Exactly. Find the one spot screaming for help, measure it obsessively, then earn the right to add more layers. Don't boil the ocean. That's how you end up with a data nightmare. Rachel: And here's the operator thing that keeps me up: the platforms Shopify, Amazon, Cart.com, they're controlling your data and deciding where your AI lives. That's where the real leverage shift happens. Max: Yeah, because when the CFO says thousands more merchants go live on agents in weeks, that's not just a feature drop, that's real platform leverage and operator risk you need to think about. Rachel: Right, so, when agentic commerce scales, who actually wins—the merchants or the platforms holding all the data? Max: Spoiler alert: if you're not thinking hard about where your AI lives and who owns that intelligence, you're already behind the game. So we teed this up last segment, but now we got to talk power players. Who actually owns your customer in 2026? And more importantly, who owns the data? Rachel: Here we go, the custody battle. Max: Exactly. And Shopify fired a shot. Their CFO basically said thousands more merchants live on agentic applications in weeks, not years. That's not hype. That's roadmap. Rachel: Yeah, that weeks line jumped out at me. I've run this at scale. For a merchant that's not R&D, that's forced roadmap. Shopify is making a bet and pushing it down whether you've built the pipes to handle it or not? Max: Right. If you're on Shopify, the default fast becomes agents running search, merchandising, cart building, whether you planned a massive AI project or not. It's coming to everyone. Rachel: And then their president claims agentic commerce expands the TAM? I'm skeptical of TAM expansion claims without hard metrics. Tricks. But how do you read it? Is that real incremental demand or shuffling existing pie? Max: To me, it means they're betting these agents pull in shoppers who never bothered with clunky websites. Think order olive oil with flour on your hands as normal behavior. Rachel: So basically, Shopify's pitch is let us own the agent layer. We'll expand your TAM. And the unspoken part, we own the relationship. Max: Yep, and here's the trade: you feed them behavioral data so the AI learns faster, but whoever owns that learning owns the edge. Rachel: That's the part that keeps me up. Who owns that data asset? Because I've watched this play before at Walmart. Once the platform knows everything about your buyer, price sensitivity, reorder patterns, that becomes their leverage, not your moat. Max: Totally. That's the operator question I've been obsessing over. If that intelligence lives with Shopify, you're not owning the customer, you're renting access to a profile from the platform. Rachel: And in a down turn, whoever owns the relationship dictates margin. I've literally sat in those calls at Walmart. It gets brutal fast. Max: Speaking of brutal, Amazon—during that big outage—ads still charged while conversions tanked. You paid for the outage. Rachel: Yeah, that one made every operator I know furious. You're paying the tech tax even when their infrastructure is burning down. Max: Exactly, and it shows the shift. Amazon and Walmart aren't just price competitors anymore, they're tech platforms monetizing every layer: ads, data, logistics, AI. Rachel: So if Shopify is playing we'll grow the pie with agents, Amazon's playing we'll tax every transaction that moves through our rails. Platform risk 101. Max: And they both sound reasonable until you look at your P&L. Then you realize, why is my contribution margin evaporating while they're making record revenue? Rachel: If I'm a mid-market brand, I need to be ruthless about this. Which channels do I own? Which am I renting? And where do I actually have leverage on margin? Max: Good question. I'd split it. Use Amazon as a demand faucet and liquidation channel. Assume you do not own the relationship there. It's rented. Rachel: Mhm. Max: On Shopify, I'd flip it. Lean hard into Agents, but with guardrails. First-party data stays synced into your warehouse and CRM. The Agent optimizes, but you keep the memory and the relationship. Rachel: So let Shopify's AI personalize, but you still require email, SMS, maybe loyalty ID on key steps so you can talk to that customer off-platform. Max: Exactly. Measure what percent of agent-driven orders are tied to identifiable customers you own. If it's all black box, You're just rebuilding the Amazon problem with better branding. Rachel: Cart.com raises $180 million saying they're the operator across channels. But I'm asking, are they reducing platform risk or just adding another layer that claims a piece of your customer? Max: They're kind of infrastructure Switzerland, sitting under Shopify, Amazon, maybe Walmart saying, we'll run your ops, analytics, and storefront as a service. Sounds clean until it's not. Rachel: I get the appeal, but I've seen that movie. Mid-market brands end up with three platforms all claiming pieces of the same customer. That's fragmentation, not leverage. Max: Yeah, and suddenly your data looks like a divorce custody spreadsheet. Rachel: Exactly. Max: From a CRO and risk lens, I'd ask three hard questions before adding someone like Cartcom. One, does this reduce platform risk or just compound it with another dependency? Rachel: Two, do I actually get richer data back or just prettier dashboards that hide the same black box I couldn't act on before? Max: And three, does this partner help me move customers toward channels I own, like my Shopify storefront and email, or keep them in marketplaces where I'm permanently renting them? Rachel: I've seen brands get torched going Shopify-heavy. One fee change, one algorithm shift, and the entire margin model evaporates. I know operators who watched it happen. Max: Totally fair. I'm not saying nuke Amazon and go all-in on Shopify. I'm saying be ruthless and explicit which channels are profit, which are discovery, which are pure data engines. Rachel: Ooh, say that again. Max: Profit, discovery, data. Amazon might be discovery plus some profit. Shopify with agents should be profit plus data. Cartcom should earn their keep on ops leverage and incremental discovery. Rachel: And if a partner doesn't clearly win in one of those three buckets, they're just margin bleed; they're a tech tax pretending to be a solution. Max: Exactly. Rachel: So for twenty twenty six. Who owns your customer comes down to this. Marketplace controls the shelf, platform controls the rails, and you either own the data and relationship or you're renting forever. Max: Your job is simple. Make sure every shiny AI feature, every new operator ultimately pulls customers toward relationships you own, not deeper into channels where you're paying tolls forever. Rachel: Which sets up what comes next, because here's the operator truth. You can own the relationship and the data. But if your checkout burns two seconds loading and your page speed is trash, you're just automating abandonment. Max: Yeah, next up, we're talking about where you actually lose the money, page speed, abandonment numbers, and how payments and AI offers claw that back. Rachel: Stay with us. This is where the ROI math actually shows up on the P&L. Your CFO is going to want this. Max: If Shopify and Amazon are fighting over agents, the real battlefield is still your checkout button, because that's where the actual margin shows up. Rachel: Yeah, this is where margins get made or burned. It's the ultimate ROI test, and I've watched both happen in real time. Max: That one's sneaky, and honestly brilliant. Honestly, just build guardrails. Don't upsell if they're on slow networks and already hesitating. Keep checkout lean. Let the agent game come later. Rachel: So more guardrails AI than let's add five carousels of recommendations? Max: Exactly. Rachel: Let's talk KPIs, because finance does not care that you feel faster. Max: No, they do not. Rachel: If I'm walking into a QBR, what's the KPI stack I bring that... that actually speaks finance language and moves gross margin per order. Max: I'd bring four, and these are operator metrics, not just vanity numbers. One, Checkout conversion rate. Two, time to complete. Median and 95th percentile, so you see where people get stuck. Three, payment failure rate by method. Four, cost per successful order. Fees plus fraud divided by orders. That's the one that makes finance lean in. Rachel: I love that last one. That's the bridge from we added wallets to We improved gross margin per order. Max: Exactly. And it bridges technical work straight to gross margin. That's the language operators speak. Rachel: Give me something I can test, measure, and prove ROI in 30 days. No six month roadmap theater. Max: Three tests. Test one, Speed AB. Strip one variant down to the absolute minimum. No upsells, no trust badge overload. See if conversion lifts when it's just lean and fast. Rachel: So literally, stripped down and fast beats pretty and slow. Max: Exactly. Test two: wallet prominence. Take your top express method (usually Shop Pay or Apple Pay) and put it above the card form for half your users. Track usage, conversion, and time to complete. Rachel: If time to complete drops thirty percent and conversion lifts three to five percent, that's huge! Max: Yep. Third test: method mix by device. Mobile gets wallets first. Desktop gets more card and BNPL. L. Compare failure rates and abandonment by cohort (you're teaching the algorithm your best rails by context). Rachel: And all three, you can build, run, and read results in 30 days? That's real operator pace, not consultant pace. Max: Exactly. All three you can build, run, and have clean results in 30 days. That's how fast you should be moving on this. Rachel: I like that these also prep you quietly for Agents. If your Checkout is fast, supports modern rails and you understand your cost per successful order, you're ready for machine buyers. Max: Yeah, your agent ready before agents even show up. You've built the foundation, fast rails, clean data, measurable unit economics. That's the unsexy stuff that wins. Rachel: And of course, once the order is placed, everything breaks if your ops can't execute cleanly. That's the unsexy infrastructure work that actually determines whether you keep the customer or burn margin. Max: Totally. Next, let's get into it, how warehouses, carriers, and logistics are quietly rewiring for agentic demand, because none of this matters if fulfillment breaks. Rachel: Your data, your inventory sync, your fulfillment readiness, fix that before you turn on agents. Otherwise, you're just automating chaos. Max: Checkouts where you win the order. Ops is where you keep the customer. That's the real game. So if checkouts where you win the order, ops is where you actually win the year. That's where margin lives or dies. Rachel: Yeah, this is where you either make margin or burn it. There's no middle ground. Max: Exactly. And you can see it in who's buying whom right now. Barrett Distribution snapping up last mile outfits. Austrian Post buying EU shipments. That's not coincidence. That's consolidation around where the volume actually lives. Rachel: Because like unsexy infrastructure is exactly where I've watched real leverage hide. Max: Not at all. But it screams one thing. Omnichannel and cross-border are where the money's going. The boring infrastructure, the plumbing nobody wants to talk about, is where real leverage sits. Rachel: Right. Customers want frictionless. Brands want one partner instead of a margin-destroying spreadsheet empire. Three customs lawyers. Max: Exactly. And the folks who own that plumbing see demand patterns across hundreds of merchants. That's when AI stops being a chatbot toy and becomes a margin driver. Rachel: Okay, go there. Because earlier we talked agents on the front end helping you buy things. What's the ops version of that? Max: I used to spend literal days, full days buried in spreadsheets just trying to answer, are we over or understocked? That's what AI was built for, pattern recognition at scale on boring problems. Rachel: I'm with you. But operations first, then AI. I learned that the hard way. Bad inventory data doesn't get smarter with a replenishment bot. It gets faster at burning money. Max: Totally. This is the same rule. No fancy ops agent before you fix your data foundation. Get the inventory house clean first. Rachel: Exactly. Three non-negotiable hygiene steps before anyone even whispers about automation. Max: Let's do it. Rachel: Okay. One. SKU hygiene. One ID per product per location. No alias SKUs. No red T final final dot V2. I've debugged this at scale. If your store, WMS, and 3PL all call it something different, AI double orders or never orders. Max: Right. If your team can't manually reconcile it, the model's going to be worse. Garbage in. Garbage out every time. Rachel: Two. (Inventory Truth) Pick a source of truth and sync. Amazon says ten, Shopify says three, warehouse says seven-I've literally seen teams try AI reordering on that mess. The stockouts got worse, not better. Max: Yeah, that'll do it. Rachel: Three. Basic demand tagging, seasonality flags, promo tags, product life cycle. If AI thinks Black Friday is the new normal, you'll be overstocked for six months and watching that margin evaporate. Max: Yeah, the machine's telling you growth is crushing it while your warehouse is literally drowning in returns and overstock. Rachel: Exactly. Get those three right and the agents actually have clean signal to work with instead of garbage patterns. Max: And that's where this gets fun. Imagine 18 to 24 months from now, the default eCommerce stack is just running itself, automated end-to-end. Rachel: Paint it. Max: Okay, front-end, AI search, recs, offers, checkout fastest rail picked, fraud screened, payment routed to the cheapest processor, all real time. That's the full picture. Backend: An inventory agent watching sales, lead times, margins, auto-adjusting reorder points per SKU per region. A logistics agent picking the optimal carrier each order. Cost, speed, risk, all balanced automatically. Rachel: And maybe a returns agent quietly sandbagging promos on products that come back more than they sell. Max: Yes, plus finance gets a margin co-pilot that tells you push that free shipping banner and here's your contribution profit hit given current carrier rates. That's real operational intelligence. Rachel: That's the dream. But real talk, you don't need a million dollar platform. You need to prove it with your own data first. Max: One hundred per cent. Rachel: So here's one real experiment: next week, small, measurable. Max: Let's go! Rachel: Take your top twenty SKUs by revenue, write it down, on hand, last thirty days of sell through, restock lead time, whether it's perishable or high return. No vendor theater, just the numbers that matter. Max: Pen and paper, good old spreadsheets, the foundation before the robots. Rachel: Yeah, and then ask where would my reorder logic change if this was automated? Faster pulls on long lead time items, smaller frequent orders on perishables, where expiry kills the unit economics? Max: Exactly. You're basically writing the playbook an agent would automate later. Manual first, then scale. Rachel: Exactly. And if that manual tweak moves your KPIs, stockouts drop, waste tightens, you've proven the ROI case with your own data. That's the argument finance actually listens to. Max: And you flush out the data mess on 20 SKUs before you turn on automation across 1,000. That's how you avoid the disaster. Rachel: Yep; start Narrow, get it clean, then let the robots loose, not the other way around. Max: So to tie a bow on the whole episode, AI is rewriting the funnel, checkout, and now the back office infrastructure-the entire game. Rachel: But the real Edge, it's not the flashiest model. It's the operator who cleans their foundation then points AI at the problems that actually bleed margin on the P and L. Max: Speed up the click, clean up the stack, let the agents handle the small stuff, while you focus on margins. Rachel: Try that little inventory exercise this week. See what you learn, and we'll see you at the next Checkout Point. Max: Later. All right. That is the Checkout Point for today. I'm still stuck on that Berlin Expo bit. I need a week of outfits for rainy meetings and the agent just build your cart. That's Checkout done right, man. Speaker 3: Wow. Rachel: That visual is going to live rent free in my head. And if you remember nothing else, Pick one AI wedge, measure it hard and prove ROI. That's the operator playbook. Max: Exactly. Don't boil the ocean. Fix the product grid. Search. or Checkout first, then earn your way to Agentic magic. Rachel: And Max: And Please, for my sanity, track the KPIs we talked about so you're not just burning margin on someone's AI theater. Rachel: Yeah, yeah, if your AI orders go 15x and total orders don't budge, the shells are moving, not the money. Max: So if this was helpful, hit subscribe, leave a quick review, and share it with the one person on your team who always says, we should try this AI thing. They might be your next operator. Rachel: We actually read those reviews, track what's landing, and it helps get this in front of more operators who need to hear it. Max: Thanks for hanging out with us today. Rachel: Next week, we're going ops and logistics, all the unsexy, boring stuff where your actual margin lives. Max: Until then, keep your carts fast, your agents honest, and your KPIs sharp.