Subscribe
Share
Share
Embed
10 minutes of expert insights every weekday. Your morning ritual for staying ahead in retail media.
OpenAI Killed Instant Checkout. But Don't dance on the grave of agentic shopping yet
===
[00:00:00] Kiri Masters: Shortly after OpenAI Instant Checkout launched last year, I made a fully Agentic purchase through chat GPT. It was a lavender candle. The experience was fine, not bad, not [00:00:15] broken, just not meaningfully better than what I already do on Amazon or Target. I never went back. Now, if I. As someone who covers this space for a living and genuinely wants AI to [00:00:30] handle more of my shopping, couldn't be bothered to return.
[00:00:33] Imagine how the average consumer felt. So when the news broke a couple of weeks ago, that open AI is killing instant checkout and routing purchases back through [00:00:45] retailer apps. I wasn't exactly. Shocked, and I also wasn't shocked by the victory lapse everywhere. AG agentic commerce is dead. We told you so the naysayers were having a field day, but I reckon they're [00:01:00] celebrating the wrong thing.
[00:01:01] Let's jump in.
[00:01:02]
[00:01:04] Kiri Masters: there's no chance that you miss this news by now that OpenAI is deprioritizing instant checkout, which is the feature that let people buy things through chat GPT. [00:01:15] Going forward, if you want to actually purchase something within chat GPT, you'll either need to use the inbuilt app inside chat, CBT, like what we have for Instacart or Expedia right now.
[00:01:29] Or you have [00:01:30] to bounce over to the retailer's site to complete the transaction there. As it turns out, retail is complicated, product data is messy, unstandardized fragmented. This is all really, really [00:01:45] hard.
[00:01:45] I read a great essay in a newsletter from Aura Labs, which posited that part of the reason for this all breaking down was that OpenAI had tried to be both the pipe and the toll booth inserting [00:02:00] itself as a new checkout intermediary and taxing every transaction.
[00:02:04] Users rejected the toll booth, .
[00:02:06] The market was right to kill it, but the commentary has made a category error. It's treating one failed implementation [00:02:15] as proof that the entire behavioral shift is a mirage, because while OpenAI was fumbling checkout, agentic shopping is already working in another market.
[00:02:26] Alibaba's Quinn app, which [00:02:30] is. An AI assistant completes food orders, travel bookings, and product purchases inside a single conversational interface right now at scale. Now it works because Alibaba owns the AI [00:02:45] model. It owns the marketplace, it owns the payment rails, it owns the logistics. And as this essay from Aura Labs points out, OpenAI tried to replicate all of this without owning any of those things.
[00:02:58] So that is what has [00:03:00] failed, not the concept. So we know that this behavior can work. The question is how and when it arrives in the west. Now before the Quinn comparison gets anyone too excited, not every behavior that works in China [00:03:15] catches on in the US, or at least not on any predictable schedule. Let's think back to live stream shopping.
[00:03:23] It was the big thing in China around 2016. It grew into a huge market. In [00:03:30] recent years, but in the US the first wave of this implementation flopped meta shut down Facebook live shopping in 2022, Amazon launched its TikTok, like short form video and [00:03:45] Photoshopping Feed Inspire. Remember that? Maybe not. It didn't last very long.
[00:03:51] The reasonable conclusion is that Americans don't wanna shop this way, and that store lasted years and years, [00:04:00] and then TikTok shop launched. In September, 2023 and found traction by embedding Commerce inside Entertainment, create a trust native checkout. Now, these things had never [00:04:15] really all come together before, but it worked.
[00:04:18] when it became a natural extension of an app that people already loved. Now there is a two, a double-ended lesson here for the naysayers. It a [00:04:30] stall doesn't mean that the behavior is dead. And for the enthusiasts, myself included. Just because Quinn proves that the model works in China doesn't mean it'll translate directly to Western markets on any predictable schedule.
[00:04:44] People don't [00:04:45] adopt new technology because it's theoretically useful. They adopt it when the experience is natural, it's worthwhile and it's safe. Perhaps the chat GBT version of a gentech [00:05:00] checkout didn't check those boxes just yet.
[00:05:03] That's an execution failure, not a conceptual one.
[00:05:07] I've spent the last few months arguing that AI enabled commerce poses a threat to the $60 billion retail [00:05:15] media industry. That when discovery moves upstream into AI assistance, that onsite advertising surfaces, where retailers earn 70 to 80% profit margins start losing their audience. Nothing [00:05:30] about the news from a couple of weeks ago changes that thesis discovery has already moved upstream. People are using AI assistance
[00:05:38] for discovery, for research, for comparisons. Amazon isn't pulling back. They have [00:05:45] everything needed to make agent shopping work. They have the user context, they have the product graph. They have the payment rails, they have the F fulfillment infrastructure.
[00:05:55] Rufuss is getting better and better all the time. Now this is [00:06:00] starting to sound a lot like that Alibaba Quinn comparison. Which is the same. Alibaba owns all of those rails. So Amazon, as it stands, is the closest Western equivalent to what is happening there. [00:06:15] Amazon is also signaling interest openly in enabling purchases from other retailers through their AI tools, through buy for me.
[00:06:25] Or shop direct. I think it's undergone a brand change that is the [00:06:30] real competitive threat. Another threat is Google. Google isn't sitting still either. They have the shopping graph, they have the merchant center. They're pushing Gemini and AI mode aggressively. Just ask my mom, [00:06:45] she. Is not a fan of Gemini everywhere on her Android phone.
[00:06:50] Google has fumbled in commerce before, but the pace of recent developments suggests a willingness to disrupt themselves that we haven't seen from them in some time. [00:07:00] Miracle Ads is the only retail media solution designed for both one P and three P Marketplace brands. Why does that matter? [00:07:15] Marketplace sellers demand a seamless advertiser experience that still offers full funnel ad formats, and retailers need a flexible solution that allows you to scale your media business.
[00:07:28] Learn [00:07:30] more@miracle.com. That's M-I-R-A-K l.com.
[00:07:36] Now here is a speculation that has been posited from a number of people. Maybe that big investment that Amazon made in [00:07:45] OpenAI recently could have something to do with this pullback that is complete conjecture.
[00:07:50] It's a very interesting idea to consider. Could be true. We won't know until when or if something either validate, validates [00:08:00] that assumption or not. We'll have to wait and see. Wrapping up here, look, the bar for meaningful change in consumer behavior is very high. My own experience with the lavender candle confirms that [00:08:15] one fully agent purchase never to return again.
[00:08:19] But here's what's happened since I made that first underwhelming checkout purchase. AI has become indispensable to every other stage [00:08:30] of my buying journey. I use it constantly for product research, comparison, finding deals, figuring out what I need in the first place. That first 90% of the shopping journey has already been rewired, at least for [00:08:45] me, and and data from many studies suggest that I'm not alone the last 10%, the actual transaction.
[00:08:54] That is the hard part. There are many, many technical hurdles to overcome order management, product [00:09:00] feeds, payment safety, et cetera. It also needs to be appealing to consumers. What about loyalty program in integration? What about handling returns and refunds? What about smart replenishment? That is a [00:09:15] better experience than what we have with subscribe and save, cracking those things that would provide a genuinely additive consumer experience.
[00:09:25] Those are also very difficult. One last comparison that I like to [00:09:30] make. Consider how Amazon's own one click ordering played out when it launched all the way back in 1997. Buying something with a single button click felt risky to a lot of [00:09:45] people. A lot of people had never even bought anything online before.
[00:09:48] There's no confirmation pages, no review step, but it worked because. Eventually shoppers trusted the platform enough to let go of a familiar friction point. [00:10:00] Amazon patented it. Defended their patent aggressively, and it became one of the most consequential UX innovations in e-commerce history. The Buy Now button of the agentic future will [00:10:15] need that same combination, a trusted platform.
[00:10:19] A solved operational backend and an experience that makes the old way feel unnecessarily clunky. By the time someone cracks it, we'll all be [00:10:30] so embedded in an AI assisted shopping path at every other stage of the journey, that the final step will feel like the obvious missing piece rather than a leap of faith.
[00:10:41] The skeptics got the short term call, right? [00:10:45] This particular version of ag agentic commerce was oversold and underbuilt. But are they right about the destination or, or the timeline? The grave dancing feels premature to me.
[00:10:57] I'd keep your running shoes [00:11:00] on.
[00:11:00]