5 Agentic Commerce News Items: Big Moves by ChatGPT, Meta, Amazon and Two Sets of Fresh Agentic Commerce Data!
We've been holding 5 news items for a super-post and if you're going to Shoptalk, we want you to be the most informed people there. See you soon!
Retailgentic at Shoptalk Reminder
Two quick reminders:
Agentic Commerce Debate - I’m speaking at Shoptalk Tuesday March 24th at noon in the big Agentic Commerce debate - Agentic believers come cheer on Ekta Gupta and I - Naysayers - don’t miss lunch, you’ll want to head to the cateteria early 😄
ReFiBuy booth - ReFiBuy’s booth is SU3 - it’s in the startup area by the food/eating area in the exhibitor hall.
News Item 1: ChatGPT’s Checkout Confusion
I’ve received many requests from readers in comments, LinkedIn and DMs asking about my thoughts and predictions on the controversy around ChatGPT’s instant checkout changes/confusion. All I can say about the why is that is I have some conflicts I have to manage through.
There’s enough public information available , that I think you can form your own opinions and make a good prediction of what’s REALLY going on if you look at it through the right lens and the right information thats been presented. Here’s my assembly of the information presented and you can form your own opinions.
Information Asymmetry in the Ecommerce Industry
One general observation before we dig in here, I’ve talked to hundreds of reporters and folks outside the industry about ecommerce over the last 30yrs. Similar to the disparity between the consumer UX (so simple and easy!) and ecommerce back-end implement details (insane complexity required to deliver that simple front-end UX), it is extremely hard to get nuanced, complex information about ecommerce across to people not building in the space. Also complex, multi-faceted details aren’t really great for sound bytes and awesome headlines. Can you imagine this headline: “MerchantX says if only the canonicalization, variation matrix and supplemental contextual+conversational datafeeds were easy to implement with the identity linked loyalty programs, it would be easier to do X, Y and Z” - sadly not so riveting when you get into the weeds.
Timeline Refresher
9/29/25 - OpenAI ChatGPT announces Instant Checkout
11/30/25 - Walmart launches on Instant Checkout
1/25/26 - Retailgentic discovers ChatGPT multi-item/single-merchant cart system here and on 1/28/26 more details here.
2/8/26 - Retailgentic discovers the ChatGPT self-service merchent center here.
2/22/26 - Retailgentic discovers 2 more features for ChatGPT instant-checkout: more info and ‘follow-up’. detailed here.
3/4/26 - Information runs this piece: “OpenAI Scales Back Shopping Plans for ChatGPT”
3/18/26 - ChatGPT’s head of ChatGPT, Nick Turley, when asked a direct question about Instant checkout dodges the question and says they are focusing on discovery right now. Our coverage here.
3/19/26 - Google announces UCP expansion with: Multi-item cart, improved discovery endpoints, loyalty integration and
3/4/26-3/22/26 - Headlines about Instant checkout and the death of Agentic Commerce flood the zone as the Naysayers pile on and dance on the grave of Agentic Commerce.
My three favorite examples of the ChatGPT Checkout/Agentic commerce 🪦🕺
Why Walmart is Moving from Instant Checkout to App Checkout?
The Tech Buzz article here and a Wired article out here, what we have presented from these two reports is a tale of Walmart not being happy with Instant checkout, surfacing problems and a ‘solution’ that raises more questions than it answers.
Instant Checkout Problems Raised (by a named source I’ll point out)→
Conversions - "Conversion rates—the percentage of users following through with a purchase of an item shown to them by ChatGPT—have been three times lower for the selection sold directly inside the chatbot than those that require clicking out, according to Daniel Danker, who oversees design and product for Walmart."
Consistency - We learned that our customers want consistency across every touchpoint," a Walmart spokesperson said in a statement.
Flop - According to Daniel Danker, “Put simply, Instant Checkout has been a flop.” Not much actionable over here, Mr. Danker is clearly not a happy camper, that much is clear.
Three other common Instant Checkout Issues I’ll throw in the mix here are:
Data errors- Frequently you’ll click on a Walmart widget A instant checkout and get widget B. This is Walmart sending a janky datafeed, that ones on them. (sorry it’s obvious).
Merchant wait times - Most merchants I talk to are very upset they filled out the ChatGPT form and never got an email/call or anything.
Crawler errors - Frequently in the discovery process, there will be mapping issues, pricing errors and incorrect shipping information. We know how to solve all this - $400B goes through Amazon’s marketplace alone and it’s the same exact type of system. These are known systems, we can build them. We, as an industry, HAVE built them. We will build more (e.g. TikTok Shops).
Reported Solution:
Next week, Walmart’s chatbot, Sparky, will begin operating within ChatGPT—essentially a chatbot inside a chatbot. A similar setup will arrive in Google’s Gemini chatbot next month."
This is a perfect example of information asymmetry - this makes zero sense to me - I look forward to seeing how the consumer going ChatGPT→Sparky→Walmart makes any sense and improves conversions. 🤔
Scot’s, aka Captain Obvious, Solution:
I’m going to just say this, it’s so insanely obvious but nobody is saying it, so I will.
The way to solve all these problems is these features:
Loyalty program integration, multi-cart, and more pdp details - Solves conversion and consistency. If I’m a Walmart+ subscriber (I am) and I can’t use that benefit on Instant Checkout, I’m going to abandon that cart. Everybody knows the economics of 1 item/1 cart don’t make any sense also. More items in a cart, more buy-in, more conversions. 🤦♂️
Self service merchant center - Once you have a self-service merchant center, the lines and forms go away AND the data quality goes up. Add a refined discovery endpoint (like UCP just did!) and bang, problem solved. The next step, right out of Amazon’s and even eBay’s 3PM playbook is you then create merchant-level incentives via strikes/scorecards in the self-service merchant center to make sure they enforce canonicalization (👀 at you Walmart datafeed) , data freshness and other key pieces to the CX. I lived 20+ years of life worried about SNAD, INAD strikes, ODR rates and the like at ChannelAdvisor.
Dang, if only ChatGPT had these features 🧐
The only piece we didn’t solve in the marketplace era was loyalty programs and merchant-of-record. Instant Checkout solved MOR first, and it brings with it the unlock to loyalty. You’re one inch from loyalty when you solve MOR which IMHO is the harder problem.
I will say this as clearly as possible. These are all solved problems (Google UCP now has gone that last inch, so I’m counting it as ‘solved’ now too btw).
Ironically, Walmart knows exactly how to solve these problems (except MOR/loyalty), because Walmart solves them today right there it’s own third-party marketplace.
Analysis of the Information Piece
From the information piece, what we have at the core of that piece is this quote:
“We’re evolving how we approach commerce in ChatGPT to better meet merchants and users where they are,” an OpenAI spokesperson said in a statement, adding that the company is prioritizing improving product search and discovery in ChatGPT. “Instant Checkout is moving to Apps, where purchases can happen more seamlessly. We appreciate our partners for learning with us and look forward to sharing more as we continue building in this area,” the spokesperson said.
A couple of things to point out:
The quote is attributed to an un-named “OpenAI spokesperson”
The mystery person explicitly says: “…we continue building in this area.”
Unfortunately, we don’t know if the reporters at Information asked these critical follow-up questions:
“What about the multi-cart experiment”
“What’s going on with the merchant center?”
"How about the expanded experiences we’re seeing being tested such as more info and follow-up?
“Any update on loyalty program support?”
“Is this just a Walmart thing or an everyone thing?”
“Ok, you are going to focus on discovery, are you going to stop checkout all together?”
The Instant Checkout Dog Not Barking…
Editorial Note: This is an idiom based on a classic Sherlock Holmes story. Details here for my many international readers that I may have lost on this reference. 👍
Swirl all this together and do you see the dog not barking yet? Let’s look at it from 30k-feet:
I. Instantly Checkout 1.0 is obviously just a first step
Literally everybody in the industry knows that Q4_25 Instant Checkout wasn’t complete. In this June 2025 articleutm_source, I introduced the Five levels of agentic commerce and very purposely put ‘loyalty program integration’ as part of ‘Advanced buy’ because it and multi-cart are obvious ‘must haves’ for this to work at all.
II. ChatGPT is clearly building features to address IC 1.0 shortcomings
We have very clear and transparent evidence in Q1 2026 of ChatGPT building out multi-item cart, a self-service merchant center and mini-PDP enhancements for the product and offer cards to improve conversion by giving the shopper much more product information.
III. Google announces UCP enhancements: Loyalty, multi-cart, discovery and self-service merchant center.
The UCP team at Google, as surfaced in my interview last week with Ashish, has clearly seen the need for these things from the three stakeholders of UCP partners, shoppers and the sea of merchants interested in these programs.
I hope this presentation of information has enabled you to see the dog not barking here and helps you start asking some deeper questions about what’s really going on here.
News Item 2: ChatGPT's Shopping Home Page Take II
Speaking of ChatGPT and discovery, Joe the Agentic Sleuth has an update on the new ChatGPT Shopping Home Page. This one is best seen as a very short < 30s video clip here→
Here you see that since our first reporting way back on March 18th (here if you want to compare before/after) it’s already much cleaner and clicking on the discovery cards actually fires a prompt and shows relevant results now.
I know what you’re thinking - kind of crazy that they are still working on this disaster of a flop, what gives?! 🤪
News Item 3: Meta Decides to Go ACP?
Several folks that watch the ACP Github repositories like crazy have told me of evidence in that repository that META is now an active ACP participant, but so far, no evidence of them in UCP. Several of these folks have basically come to the somewhat logical conclusion from this chain:
Jan 29 - On their Q4_26 Earnings call Zuck said they are doing Agentic Commerce. Our detailed coverage here. No direction on UCP/ACP mentioned.
March 2- We discovered that Meta AI is Live with Agentic Commerce results, but no checkout - yet.
March 15-18 - Reports of ACP activity by Meta go from zero to low roar
At this point if you are getting in the Agentic Commerce game, you have two, or you could call it three options:
Option 1: Adopt existing protocol:
Option 1a. Adopt existing ACP
Option 2b. Adopt existing UCP
Option 2: Build your own new protocol
It seems kind of obvious to me given they are so strong on the discovery side with Instagram and all the surfaces, they would go Option 1. At the same time, Anthropic, being anthropic I believe will go Option 2 and when they do it will blow our minds. They have an uncanny ability to jump 18 months into the future. I suspect time travel is involved and Dario is from 2040, I could be wrong that one, this is not an official Retailgentic prediction.
Conclusion: I think we’re at 75.4% probability we get a ACP announcement out of Meta AI in the not too distant future. I’ll point out the math on their next earnings call would put it in late July. I think that’s the ceiling on something - that’s like 5yrs in AI time.
News Item 4: McKinsey Reveals Feb 2026 Consumer Agentic Commerce Preferences
We haven’t had a great, reputable n=2000+ consumer since November of 2025 and the long wait is over. McKinsey has a great research paper here and they are at n=4008 - those last 8 - 👍
There’s a TON of great info in here and I recommend you go check out the whole thing, the two pieces of data that caught my attention as newsworthy and worth discussing at cocktail parties, board rooms and Shoptalk are:
Datapoint 1: What Shopping Activities are AI Consumers Engaging In?
I’ll note while trying not to say any more snarky comments about the first topic that 23% are now using it to checkout and buy. Wow. I was told it’s a mirage AND a hallucination, I guess it’s heading towards the ‘Mass Delusion’ phase. (Ok, to my credit that was snarky about naysayers generally, I’m warming up for Tuesday at noon).
Datapoint 2: What Categories Are AI Shoppers Shopping?
This is interesting, the last data from November had a bit different sequence with grocery very high (31% in McKinsey, and 49% in MS AlphaWise below). Note that McKinsey has Electronics at 38% - by far the highest i’ve seen that and AlphaWise is at 25%.
News Item 5: Is Amazon Working on an Alexa+ Phone?
I’m pretty sure that Jason RetailGeek and I are the only two people holding on to our vintage Amazon Fire phones (the first, and last 3D phone!) The Fire phone came out in June of 2014 and a short 14 months in August of 2015, it was cancelled.
Imagine everyone’s surprise after that famous debacle to read the exclusive reporting by my OG buddy Greg Bensinger, currently at Reuters who in this piece, wrote:
Amazon’s new phone project is codenamed ‘Transformer’
Focus on AI integration, Alexa features, and mobile personalization
Led by Amazon’s ZeroOne group, inspired by minimalist Light Phone
Four people are sourced, but not named since they are not approved to speak publicly about it.
Friend of Retailgentic, Wall St. analyst Colin Sebastian had three interesting points in his analysis→
What do you think - can Amazon nail an AI device and create the ‘next platform” or will the mystery device be a repeat of the Fire phone?
News item 6: LogicBroker Releases Industry Data.
This one is a bonus because it’s for the fellow industry people that read this that are building agents and companies in the same ecosystem as what we’re building at Retailgentic. I live in this weird world, where on LinkedIn, X and the press, I’m seeing constant messaging about how Agentic Commerce is a dead end, will never work - you now the drill. Then in the day job/reality-land, we’re bombarded by interest from brands and retailers across the globe having a ‘hair on fire moment’ trying to figure out what’s going on.
Then you see n=4008 consumer survey data that 23% of people are using Agentic Commerce to checkout. Make it all make sense?!
This is a survey that is n=600 which for consumer is not enough, but this is 600 executives at Brands and Retailers, which is a significantly smaller TAM than ‘consumers’, so it’s actually a large sample set from what I’ve seen historically.
Founded back in 2010, LogicBroker started as a dropship network and has added oms/order routing type capabilities and is now touting AI capabilities. I mention this because it would be unusual for them not to ask a lot of questions about this aspect of AI and merchants.
Three charts to call out:
Chart 1: This Report Covers Retailers of all Sizes:
I always check these industry surveys for merchant size - there are millions of online merchants, most are < $1m/yr - this report has a good mix of large and medium, and not a lot of the smaller merchants.
Chart 2: Merchant Leadership Belief and Investment in Agentic Commerce
These three stats jump off the page:
91% expect AI to influence at least 20% by 2027.
57% are implementing agents
47% are investing over $1m in 2026
If you are, like me, in the enablement biz, this helps you prove to the team and potential investors out there that this not a ‘nice to have’, it’s real, with real budgets and real projects that are getting funded.
If you are a leader in retail/brands, you may feel like you’re on the cutting edge and alone, but you are not, we are at a tipping point of 50% and in another 6 months I suspect we’ll be into the 70% zone. This pairs nicely with the McKinsey report showing strong consumer adoption, it’s logical that smart retailers and brands will follow their data and come to similar conclusions.
Chart 3: What Areas of Investment Are Leaders Focused On?
You can see the top two are:
AI Powered Product Discovery - They don’t specify if this is off-site or on-site, I believe in my conversations with retailers, they are pausing replatforming and a lot of new on-site work to invest in the new world of off-site discovery. This leads at 50%
AI Chatbots for service- These agents are in the customer service function and helping drive a lot of economic value for retailers and faster/better answers to shoppers. A close second at 48.5%, but Discovery edged otu the win (Yeah discovery!)
Those are a couple of tidbits, the whole report is really excellent - thanks to Logicbroker for doing this.
The full report is here and is sourced as: “Logicbroker/Midsail Research “The State of Agentic Commerce Adoption.” 2026: 2026 Market Research Report.
That’s a Wrap!
Unless there’s really big news, that’s it for this upcoming week, we’ll be back in full Retailgentic force next week and hope to get some Shoptalk highlights out while they’re fresh!