AI agents are about to kill your business

This This is the Billiondollar Sellers podcast. Your go-to source for cutting edge strategies and success stories from the world of Amazon and e-commerce. Buckle up and get ready to take your Amazon business to new heights. Don't forget to subscribe to the Billiondoll Sellers Newsletter. Welcome your host. Welcome your host, Kevin King. >> Hey everyone, welcome to the Billiondollar Sellers podcast. I'm your host Kevin King and today is April 30th, 2026. Uh we got a packed show for you today. John Aspenol did a demo inside BDSC showing how OpenAI's codeex is changing the game for Amazon listing images and A+ content. We're also going to talk about the Mega Earning Showdown where Google, Meta, Amazon, and Microsoft all reported inside a 2-hour window. Rufus just got eight brand new features that fundamentally change what a search result is on Amazon. and I'm going to walk you through every single one. Plus, uh, Chris Gray came on the marketing misfits and laid out why Google's shift to agentic AI is about to steamroll local businesses that aren't ready and some really interesting stats about who's actually selling on Amazon in Europe. But first, here's your stunt Bezos question for today. So, Amazon said yesterday on their Q1 2026 earnings call that Rufus is attracting more users, doing product research, tracking prices, and automatically buy at certain prices. And the question is, how much year-over-year do they say Rufus usage is up? Think about that and I'll give you the answer at the end of the show. All right, let's get into it. So, if you're not a billion-dollar Solar Club member, you are missing out on exclusive live demos and deep dives like this next one. And I'm not just saying that. Yesterday inside BDSC, John Aspenol showed exactly why he switched from Clyde Code to OpenAI codeex for imageheavy work. And he did it in the last 48 hours. His take was pretty simple. If you do anything visual, Codex is now the move because of the new Chad GBT image model. But if you don't touch creative, stay on cloud code. He took an Amazon listing for a kid's electric toothbrush, built an entire image strategy, generated the full image stack on brand, product accurate, no hallucinations, and it even reordered the images for better conversion. And this was done in two prompts only. It spotted that the brand wasn't using its own font and colors and corrected it automatically. It replaced those harsh red X comparisons with grayscale, which is actually better psychology on a product page. And then it generated a full A+ content premium layout in 7 minutes. John said that at his agency, that's multiple days of designer time. 7 minutes. He built brand guidelines from scratch off just a website URL like color palette, gradients, typography, tone, dos and don'ts, icons, hero examples. It built a brand store layout module by module from a website, generated YouTube thumbnails and 16x9 Tik Tok and reals assets and 9x6 2000 x2000 image stack assets all with granular pixel control. And he built consistent character sheets, you know, front side back of a character called Suzie so that every future iteration like Suzie running or Suzie at the gym stays on brand. That's wild. Now, why codeex over Claude code right now? Well, for a start, Claude doesn't have an image model. You'd have to call Nano Banana or wire in some external API and that costs about six cents per image. Plus, you get this disconnected workflow. Chad GBT Pro is $100 a month, same price as Claude Pro, and John burned only about 19% of his rate limit doing two image stacks plus an A+ project plus a full brand guidelines build. So, it doesn't hit limits the way Claude does on image work. Codeex generates notebook elev, PowerPoint, Keynote, whatever. It's got computer use built in. So you can tell it, you know, on the 24th open send it download all videos, drop them in this folder, then run this workflow, and it just does it. And automations are native. You can say, for example, make this run every 6 hours, and it converts the workflow without cron jobs or hooks or any of that stuff. Here are some quick codeex tips that John shared with us. Go to settings, personalization, and enable memories, including the one that says augment memories with screen context. That way, Codex actually learns from what's on your screen as you work. Codeex support skills, but the trigger is the dollar sign, not the forward slash. Nobody knows why, but that's how it works. And this is a cool one. Stack them. Use cloud code inside Codeex. Let Cloud Code execute and let Codeex be the QA and the final say. Best of both worlds. And run on medium intelligence 5.5 by default. Extra high plus fast each year rate limit, but it ships fast on the really labor intensive jobs. If you want to see the full replay, you can try BDSC for just $9. There's a link in the show notes. Quick break. Uh, let me tell you about something happening right now. There's a 5-day challenge going on right now called 5 days to replace your entire Amazon PPC strategy with AI systems that run themselves. It's live daily from noon to 2:45 p.m. Eastern, April 27th through May 1st. And what they're showing you is the exact AI powered PPC systems, plug-and-play agents, and campaign templates that Sophie Society uses to manage more than $4 million a month in Amazon ad spend across more than 800 brands. You can learn how to use Claude Co-work and other Agentic AI tools to take over whole roles inside your Amazon brands. Plus, you get a library of plug-andplay Claude skills and copy and paste prompts ready to use right away. and you'll see the exact AI agents and workflows that 8 figure sellers are currently using inside their brands. There's a link in the show notes to reserve your spot. All right, you guys see this? So, the Marketing Misfits podcast, me and Norm had Chris Gray on and uh man, he came with some brutal warnings and some unconventional tactics for local marketing. And the big warning he dropped is that Google is moving to an agentic AI model where search engines will actually book appointments and make purchases for customers through something called Google's universal commerce protocol. If your business runs on a basic website and organic search, you're about to get steamrolled. Here are some of the key takeaways from that episode. Your Google business profile is the new homepage. Treat it like an SEO gold mine because it now matters more than your actual website. Always have an event or offer posted on your GBP because it forces the algorithm to wait your listing above competitors. List your hours as 24/7 because AI agents recommend open businesses. Closed competitors just get skipped during off hours. Norm shared a geo targeting trick where press releases targeting cities under 350,000 population can rank you number one on Google in 24 hours. There's an Apple wallet hack using a service called Send Push to get into customer digital wallets and send three lock screen push notifications per day. And Chris broke through a gatekeeper and landed a massive local client by simply changing his caller ID name. How cool is that? Bottom line from the episode is optimize for AI agents now or get left behind. You can listen to the full episode and subscribe to the Marketing Misfits newsletter at misfits.news. Link is in the show notes. All right, let's talk about an interesting stat. So, Danxi did some research looking at more than 118,000 active sellers on Amazon Germany, which is Europe's largest Amazon marketplace, and 57%, so almost 68,500 sellers are from China. Germany itself is only second at about 23rd%. The UK is at 4.5% and then everybody else, Italy, France, Hong Kong, they're all down around 1 and a.5% each. So, even on the biggest Amazon marketplace in Europe, more than half the sellers are Chinese. And you know, if that doesn't tell you where the competition is coming from and where it's going, I don't know what does. All right, this next one is a big deal. The mega earnings Aentic ecom showdown. And so, this is from Scott Wingo over at Retail Agentic. And uh in a first ever moment, Google, Meta, Amazon, and Microsoft all reported earnings inside a 120minute window yesterday. and Agent Tech Commerce was the story. And here's what showed on Wall Street scoreboard in the after hours moves. Google was up 7% just a clean beat and raise and they added something like five targets worth of market cap which is crazy. Amazon was roughly flat. Microsoft was muted and Meta was down 7%. So they were kind of the loser on sentiment. and Amazon won the unscientific agentic commerce mention scoreboard with the most call outs across their press release, prepared remarks and Q&A. Total mentions across all four companies was 20. So let's break this down by company. Google, UCP, the Universal Commerce Protocol is winning the standards war. They just added Amazon, Meta, Microsoft, Salesforce, and Stripe to the tech council. The founding members were Shopify, Etsy, Target, Wayfair, and Google. What that means is that UCP is steamrolling OpenAI's competing ACP standard. It's basically over. Alta Beauty just launched agent commerce inside AI mode search and the Gemini app using Google as their on-site agent partner and Sephora and Macy's are next in line. They're also testing a new ad format that shows retailers selling products that AI mode organically recommends. Seller takeaway here. If you're a brand on Shopify, the rails to agentic checkout are being laid right now. Get your product data clean now. Meta Zuck went off script on shopping partnership ads run rate doubled year-over-year to $10 billion. The rolling out affiliate partnerships on Facebook and testing them on Instagram where creators tag products and earn commission. This is basically Tik Tok shops playbook. And Zuck made this bold claim that no other AI lab is building, I'm quoting here, an AI that's really good at shopping. He sees that as core to what he's calling personal super intelligence. He hinted at future monetization through commission structures or premium tiers. And the CFO Susan Lee's reaction to his four-minute monologue was basically, gosh, I almost wish we could end on that answer. So, you could tell Zach was really fired up about this. If you sell direct to consumer, the creator tagging path on Facebook and Instagram is about to get a lot more interesting. And then Amazon, Rufus is going bananas. Three fresh data points from Jasse. 115% growth in monthly active users. 400% year-over-year growth in engagement and 20% of shoppers who interact with a brand prompt convert with that brand. That's a huge number. Amazon also joined the UCP alliance and dropped eight new Rufus features right before the call, which I'm going to get into in detail in a minute. Jasse said he thinks Rufus will be the best shopping assistant anywhere, and he's super bullish on ads inside Aentic Commerce, things like sponsored prompts and AI built creative that cuts time and cost for smaller brands. The one frustrating bit though is that Jasse still describes the horizontal agents like Chad GPT and Perplexity as having bad pricing and product info. And honestly, someone on Amazon needs to hand them a six- pager because that's not the current reality. Seller takeaway. If you're not already optimizing for Rufus with A+ content, structured product data, and brand prompts, you're leaving conversion on the table. Sponsored prompts are the next PPC frontier. So, the big picture for ecom operators, ads are coming to every agentic surface. sponsored prompts on Amazon, retailer ads and AI mode commission models on Meta, creator commerce and agent commerce are starting to merge. So you've got Meta's affiliate push on one side, AI shopping agents on the other and creators are basically becoming the front door that feeds into agent-driven checkout and the investment isn't slowing down. Holiday 2026 is right around the corner. SpaceX and OpenAI and Anthropic might all go public and everybody is throwing money at this. If you saw an Amazon direct to consumer or both, Agent Tech Commerce is no longer coming. It's the quarter that just got reported. All right, now let's get into the eight new Rufus features because this is where things get really interesting. Andy Jasse dropped a number on the Q4 2025 earnings call. It shows stopped every Amazon seller cold. Rufus is now in front of 300 million active customers and is already pulling roughly 12 billion in downstream transactions. Just yesterday, he said that usage is up 400% year-over-year. And then days before the UCP announcement, Amazon quietly rolled out a flood of new Roffus features. Eight of them, seven are personalization plays and one is fully agentic. Together, they change what a search result actually is. But before we get into the features, you got to understand the shift. Two shoppers type storage bins into Amazon right now. They get different products back, same keyword, different results, and you have no idea which result your as showed up in or whether it showed up at all. Why? Because one of those shoppers told Rufus, uh, for example, I have three kids and a small apartment. And the other said, I run a minimalist home office. Amazon's own interface now confirms it. As you shop or chat with Rufus or Alexa, Amazon saves helpful information to personalize your experience live, today, and on every search. Your listing was built for a keyword. Rufus is now reading it against a person. The first feature is called scheduled actions. This is the agentic one. You tap the new plus box in Rufus and you get schedule actions with pre-built prompt pills like recommend a new coffee variety every month. Alert me to new books by my favorite authors. Set a birthday reminder with gift ideas. Save money on cleaning supplies. Now, there are two catches. Amazon will only run a scheduled action once per day. So, when you ask for an immediate in stock alerts, immediate can mean tomorrow morning, maybe. But the big deal is this is the first time Amazon has moved past responding to a query and started acting on future intent. Gift planning happens before the event. Restocking runs on a schedule. Essentials move toward automation. The system now does the work on the shopper's behalf. The second feature is persistent account memory. For all the time Rufus has been in market, it never had real access to your order history. That just changed. Rufus is now faster at finding an order than the order history page itself. You can say something like, "What was that black thing I bought in February?" and it just finds it. No specifics required. Feature number three is personalized deals. Rufus now reads your wish list browsing history and past orders. You can ask, "What are the best deals for me?" Or, "Are items on my wish list on sale?" And you get a personalized cut of deals tailored to you. Then you have custom shopping guides. This one's going to eat a lot of category top ofunnel content. You ask a highle question like, "What do I need to start a video blog?" And Rufus pieces together a kit. removes anything you already own from your purchase history and gives you a buy plan with the gaps filled in. Holy cow. Number five is easy reorder. Simple but powerful. Just say reorder dog treats or reorder everything I used to make pecan pie and Rufus loads a cart. Done. Number six, Shop Direct inside Roffus. Hard to recreate this one in the wild yet, but the signal is clear. Shop direct, which is factory to consumer, was being plugged into the Rufus chat surface. Number seven is called compare with similar. There's a new prompt pill on most product detail pages where Rufus generates a sideby-side comparison matrix against similar products. The feature itself is great, but the execution is a little hobbled right now. The Rufus gutter on a desktop PDP is roughly an inch wide, so you're scrolling sideways through a comparison table that should fit on a screen. The UI is going to need to grow up before this one lives up to its potential. And the last feature is personalized product summary. uh another PDP prompt pill where Rufus summarizes the product specifically for you drawing on what it knows about how you shop, what you've bought, and what you've browsed. So when you put all those features together, seven of them are personalization and one is aentic, but they all share the same Amazon strategy. Single directional embrace and extend UCP lets Amazon superset inventory in a pullup push model. You can shop everything on Amazon. Amazon doesn't push your data outward. Personalization features are the same. and play with shopper data. Your order history, your wish list, your past rufus conversations, your browse history. This is all Amazon exclusives. Gemini or chat GPT don't get those. Microsoft did this to Netscape back in the day. The question is whether Amazon can do it to OpenAI, Google, and that aentic shopping wave coming at the gates. Here's the one question listing audit you can run right now. Pull up your top as read the first bullet point and ask one question. Does it name a type of person, a use case, or a life context? I'm talking things like ideal for pet owners, built for small home offices, or perfect for parents managing school day chaos. Those are identity hooks. Rufus has profile types on file. Your bullet either matches one or it doesn't. If your bullet just describes a feature and stops there, Rufus is holding your listing up against its profile library and getting no match. You're invisible to the shopper your product was actually made for. The fix is not stuff more keywords. The fix is to rewrite bullets, A+ copy, and back-end signals so Rufus can map your AS to a real shopper profile, not just a search term. And uh here's a tip from BDSS Dream 100 member Amy Whis that almost nobody is running. She said, and I'm quoting, you now have access as a marketer to data about conversations that customers are actually having about their mindset. And that's huge. But play is simple. Open Rufus on one of your products and ask it why do people buy this product? What do they buy instead? what they wish it did. Who is this product for? Take Rufus' answers, paste them into Chad GPT. Ask Chad GPT to generate 20 deeper, hyperspecific questions a real buyer would ask before purchasing in your category. Then take those questions back to Rufus. You now have the literal conversations your customers are having while making the buying decision. Not a survey, not a focus group, not a guess, but the actual mindset data. Marketing has always been psychological. Until last quarter, you had to infer the psychology. Now you can read it. Loop that into bullets. A+ backend ad copy and PPC negatives. That's how you close the gap between a keyword listing and an identity listing. Okay, now before we wrap up, here's a few more hot picks for you. Amazon's Q1 2026 earnings beat Wall Street expectations. Email surface showing Amazon colluding to raise prices and remove the buy box. And eBay is testing live streaming as a service to sellers. Links to all those are in the show notes. And here's your party shot for today. You can't control how you feel, but you can always choose how you act. That's from Mel Robbins from the 5-second rule. And I think that's especially relevant right now. You can't control the pace of change, but you can choose to act on it. The sellers who are acting right now are the ones who are going to be in the best position going forward. Oh, and remember that Stump Bezos question from the beginning? Rufus usage is up year-over-year. So, how much? The answer is 400%. 400% year-over-year growth in Roffus usage. That is not a side feature. It's a fundamental shift in how people shop on Amazon. All right, that's all for today, folks. Have a great weekend, and I'll see you again on Monday. This is Kevin King signing off from the Billiondollar Sellers Podcast.