When AI Takes Over Shopping Carts

Our analysts Andrew Ruben and Nathan Feather discuss how AI shopping agents could transform how consumers discover, compare and buy products and the implications for eCommerce.

Read more insights from Morgan Stanley.

----- Transcript -----

Andrew Ruben: Welcome to Thoughts on the Market. I'm Andrew Ruben, Latin America Retail and E-commerce Analyst at Morgan Stanley.

Nathan Feather: And I'm Nathan Feather, U.S. Small and Mid-Cap Internet Analyst at Morgan Stanley.

Andrew Ruben: Today, what happens when the shopping cart starts thinking for itself and maybe even for you?

It's Monday, August 17th at 10am in New York.

As we think about trends that are driving e-commerce, which remains a share gainer within the overall retail landscape, it seems that there's a transformation that's quickly building around agentic e-commerce.

So, Nathan, I think it's timely for us to talk today as agentic seems like it could be the next catalyzer of growth and innovation within the e-commerce landscape.

Nathan Feather: How much bigger do we think agentic commerce could make the global e-commerce market?

Andrew Ruben: Global e-commerce as we see it is a nearly [$]5 trillion market today. That implies 22 percent of retail sales. The way we see over the next five years is a $7 trillion opportunity, with growth accelerating to a 9 percent compounded rate, up from about 7 percent over the past four years. And this is partly on the tailwinds from agentic.

What we see here is this broad arc of reducing friction with e-commerce over time.

Think about how easy it is now to pick up your phone, search for some inventory, click, and the goods can be here within one, two days, if not same day. That's reduction of friction that we think physical retail can't match, and the improvements of agentic commerce. Having this agent that can help you search, help you discover – that should further the e-commerce opportunity.

We think agentic alone could add about 6 percent to the five-year e-commerce addressable market, with about 20 percent of industry volumes having some material agent influence.

So, within this opportunity, Nathan, agentic isn't one size. How should investors distinguish between AI influence shopping and fully autonomous purchasing?

Nathan Feather: To your point, there's a wide different flavors that we're calling agentic commerce. And it starts really at the top of the funnel with, you know, you could go to your chatbot of choice and say, ‘I want a hiking backpack with a water bottle slot and a place to hold my keys,’ right? ‘Show me the best options in a certain price range.’

And there you're capturing the top of the funnel, but as you click in, you may bounce out to a retailer and purchase on there. Or it could go even further, and maybe you complete your entire checkout within that specific chatbot.

Now, right now what we're seeing is about half of consumers are starting the top of the funnel at least sometimes with a chatbot, but a very small portion are actually completing purchases. And so, as time evolves, we expect that funnel to widen and start to see a little bit more of this fully autonomous purchasing; although for the most part, we think it's really going to remain top of funnel and mid-funnel.

Now, adoption does look very different across regions, partially because of different consumer behaviors. Why has AI shopping gained more traction in some markets than in others?

Andrew Ruben: I think that's right, and what we see is so far to date, agentic shopping has been led by the U.S. and China. These are the two largest e-commerce markets globally, also among the highest penetration. Some data to support it: We have proprietary Morgan Stanley AlphaWise survey that show about 30 percent of China consumers shopping using AI tools over the past month. And that compares to about 12 percent in Brazil.

Now, we do see some barriers in terms of the pace of companies' innovation, but I think this is more a matter of time. The example you give of that shopping journey, that does seem like it should be applicable globally.

There is also a second barrier, and that would be trust. We do see that consumers are using AI search, using AI discovery, and as they get more comfortable with agentic, we think the use cases can increase over time. But as we see consumers today, they're comfortable with search, but not many are willing to let AI do the full end-to-end checkout.

Ultimately, as we see it, the companies will drive the innovation, but it's consumers who determine uptake.

And that raises the question of who owns the customer journey. Do retailers keep control, or do the general AI agents take the lead?

Nathan Feather: To be frank, this is one of the major unanswered questions within this market. And, you know, we can speculate, but we're not going to know for a few years. So, let's go through the potential paths here.

I think the first goes within the customer journey. Where does the customer want to check out? Who has the best experience as you go through that journey? And early on, it's retailers. They have your purchase history. They have your payment information. They have your shipping.

To your point, they're trusted. You know if you're going to shop at one of these large retailers, you're going to get what you want. And if you don't, you're going to be able to get that refunded.

And so, we think at least early on, retailers will likely keep control of that purchase journey and actually be able to innovate a lot on site. Launch on-site agents that are able to get you to the inventory they have even faster.

But retailers could gain control over time. They can shop across multiple websites. They can price match. And so, it is going to be a question over time which of these ends up taking the lead. And the economics will change as a result of that.

And Andrew, what determines whether agentic commerce ends up generating purchases that wouldn't have happened otherwise rather than simply shifting existing sales to a new channel?

Andrew Ruben: It's a good point on the economics because let's say an agentic transaction happens on a company's site. You do still have costs, and that relates to the large language model. The conversation query going back and forth, that's going to be more expensive than a traditional keyword search.

So, here's where incrementality comes in. If you're a consumer that's having this transaction on the site, we think that gives better targeting, better information, and should ultimately put the product in front of you that you want to buy. And what this translates to is incremental sales, a sale that wouldn't have happened if you only had traditional search or an experience that you couldn't match in the physical channel.

So, we do think that if the sale is incremental and those model costs eventually come down, then that's the setup for an agentic sale to be profitable. I'd also mention the advertising business. It's important for e-commerce having suppliers that will pay to be one of the product listings up front.

Our view is that if you're searching better, then you should get better discovery, and the value of that top real estate should hold. That should be more important for the supplier with better targeting, and they can pay up for that.

But there is the risk on the other side. How real do you think the risk is that external agents divert traffic and advertising dollars away from e-commerce platforms?

Nathan Feather: Well, the risk is real, and it's really dependent on the customer journey. You know, if you go to a chatbot today, you're expecting when you type in your query, you're going to get the most accurate result that they can offer. The issue with advertising is people are paying for that top slot. It's not inherently maybe the best product. It's the person who wanted to pay the most to get that top slot.

When you go to, you know, a search website, it's not necessarily the expectation, right? You know that the first few results are going to be paid, and then there's going to be organic after that. And so, from a customer side of things, there's going to be a question of whether there's the permission to see advertising within that flow.

If there's not, you could see advertising dollars get diverted, and that is a risk. If you look at large e-commerce retailers, especially marketplaces today, a majority or sometimes all of their profits actually come from the on-site advertising that exists. And so, it's something worth watching. Although we note early on, this ended up being less of a risk than people initially expected.

Now, zooming out here, we've covered a lot of ground. So, as we think about it broadly, what are the likely factors that separate the winners here? In other words, what are the capabilities that matter most as we move into an agentic world?

Andrew Ruben: Right. And to get to those capabilities, I think agentic commerce is going to improve e-commerce as a digital service. But this still surrounds the movement, the sourcing, the pricing of physical goods.

So, I believe that the rules of retail and e-commerce should still hold. That's the fundamentals of do you have the broad selection, the right inventory at the right location that can get to the right consumer? Second, the ability and willingness to innovate. That's companies that have their own agents, that have partnerships, that are developing these tools we think will be better positioned.

And then third, thinking about some complementary assets. If you're a marketplace platform with logistics, with loyalty, with financial services, this should support the positioning depending on how the customer journey evolves. Each of these factors we think will matter in an agentic world.

And then finally, what evidence should investors watch to see whether agentic commerce has moved from experimentation to a durable growth driver?

Nathan Feather: There's a couple of different factors we're looking for here, and it's important to note we're looking for leading indicators. Given agentic is still a relatively small portion of purchases, we're trying to find those things that could identify where you're going to hit inflection points. So, a few things I'd call out.

The first are company disclosures. What are the actual retailers in this industry saying about experimentation? And are the products that they're testing actually moved into production?

Second, looking at consumer surveys and whether people are starting to use these AI tools more at the top of the funnel, we think will filter down more to the bottom of the funnel over time as additional things are launched.

And last, how is your own search behavior changing? Are you starting to see you gravitate more towards an AI chatbot as you're going through your shopping journey? Oftentimes, you'll start to see the behavior start to shift, and then the dollars flow over time.

Andrew Ruben: That's it. The shopping cart may become smarter and ultimately grow faster. But for investors, the defining question remains the same: Who owns the customer journey? Nathan, thanks for speaking with me today.

Nathan Feather: Great to be here with you, Andrew.

Andrew Ruben: And thanks for listening. If you enjoy Thoughts on the Market, please leave us a review wherever you listen and share the podcast with a friend or colleague today