I'm not sold that agentic commerce takes off. Not because the tech can't do it, but because it quietly breaks the thing that actually makes retail work. Let me explain it the way I explain it to people, with a $5 chicken.
help Start with the doubt
We have had a version of this for a long time. Google Shopping has existed for years, and it never really took off. People search on Google, sure, but then they go to the merchant's own site to buy. The "search here, check out here" loop never became how people actually shop.
There's a reason, and it isn't friction. Merchants don't want the purchase to happen outside their own front door. The sale is not the prize. The relationship is the prize, and you can't manage a relationship you never get to host.
storefront What Walmart's quiet shutdown told us
Walmart turned on ChatGPT purchasing in October 2025, around 200,000 products live inside Instant Checkout. A few months later they pulled it. In March 2026 a Walmart exec said the quiet part out loud at an investor conference: purchases made inside ChatGPT converted at roughly a third the rate of people who clicked through to Walmart.com.
The stated reason was mechanical. OpenAI was scraping retailer sites to fill its product catalog, and retail inventory moves by the hour. Prices change, items sell out, promos expire. So customers got shown things that were gone or priced wrong at the worst possible moment, right at checkout, and trust broke. Fair enough. But I don't think that's the whole story.
Think about the incentives. They already paid to build it and it was running. If it were merely underwhelming, you leave it alone and it costs you nothing. Spending more effort to actively rip it out means they learned something they did not like.
security Theory 1 — abuse
Fraud or some form of gaming was happening inside the agentic channel, enough to make it not worth keeping live.
trending_down Theory 2 — no upsell (my bet)
The sales that came through were clean and small. Nobody cross-shopped. Nobody got upsold. The basket was exactly one item, every time.
My bet is the second one, and here's the tell I think proves it. Walmart didn't walk away from AI shopping. They pulled OpenAI's checkout and dropped in their own agent, Sparky, as a plugin inside both ChatGPT and Gemini. Be findable in the models, but keep the customer, the cart, and the checkout on Walmart's own rails. That isn't a retreat from agentic commerce. That's a merchant refusing to hand over the relationship. Which is the whole point, and the cleanest way I've found to show people why is the rotisserie chicken.
restaurant The rotisserie chicken
Costco sells the rotisserie chicken for $5, and they have for years. Everyone knows it's a loss leader. They lose money on the bird. So why is it parked at the very back of the store?
Because the chicken was never the point. The walk to the chicken is the point. To grab that $5 bird you pass the whole store, the seasonal aisle, the big TV, the snacks, the thing you didn't know you needed. You almost always leave with one or two extra things in the cart. Costco is happy to eat the loss on the chicken because the journey pays for it many times over.
Agentic commerce is Costco taking that chicken, wheeling it out to the parking lot, and bolting a drive-through onto it. No store owner in their right mind does that. People would roll up, pay $5 for the bird, and leave, and Costco would bleed money on every single one. The loss leader only works because of the journey you removed.
That's the quiet damage. An agent buys the one thing the customer named. There was no aisle to walk, no better brand to notice and upgrade to, no "people also bought" moment, no reminder of the thing they actually still needed. You win the transaction and you lose the basket.
smart_toy What merchants actually want
Here's the part I think is right, and it's good news, not doom. Merchants absolutely want to be found inside the large language models. That part is just the new SEO. Show up in the answer, earn the click, pull the traffic. But then they want you to land on their brand, on their site, where they get to host the relationship and build the experience.
And the experience is where AI gets genuinely great, when the merchant owns the agent instead of a neutral checkout bot owning it.
Lowe's already shows the good version
Lowe's built an AI agent into their own website, and it's genuinely great. It sees you're buying drywall and it asks, do you have mud, do you have tape? That's a cross-sell for Lowe's and a real gift for me, because now I'm not making three trips to the store. The agent remembered everything my project needs.
That's the whole thing in one example. The merchant keeps the relationship, grows the basket, and the customer comes out ahead. Nobody had to wheel the chicken into the parking lot. The AI made the store smarter instead of replacing the walk through it.
The whole room at the Fiserv event was uncertain about AI, and especially fuzzy on agentic commerce. That's the signal. The need isn't more AI, it's someone who can explain what's actually worth building and what just looks shiny.
For the day job, it reframes the agentic-commerce conversation: don't chase being the checkout, help merchants be findable in the models and brilliant once the customer lands. For Aspire and the venture side, "teach people what's actually needed" is the demand sitting in plain sight. The chicken is how you make it click in thirty seconds.