2026.09.23 | Substack | Agents | Gaetano: 2nd-Order Agentic Shopping

I have been digging through the second-order effects of Meta’s Muse over the last couple days.

The obvious beneficiaries are pretty easy to find. Meta, Shopify, Paypal, Inod etc. But I want the less obvious.

If AI agents start doing more of the searching, comparing and purchasing for us, there are going to be existing businesses with assets that become much more useful in that world. I found one small company that I think deserves a closer look.

There are plenty of issues with the existing business. Growth has gone backwards, recent financial results have been weak, and the agentic opportunity is still tiny relative to the overall company. I am treating this as a research idea for now.

But there is enough here that I think it belongs on our radar.

And I want us to get familiar with it just incase things pop off.

Let’s quickly visit the thesis!

The company is Criteo ($CRTO).

Right off the bat, I don’t love buying stocks with charts like this. The market is clearly telling us something.

But I have a thesis. I don’t know if it really makes sense or not. But that’s why we could discuss!

The basic idea here is that AI agents should become extremely good at understanding what we want, but understanding the request and choosing the best product are two different problems. I can tell ChatGPT that I need a running shoe under $180, have wide feet, run mostly pavement and care more about comfort than speed. The model can understand all of that immediately. It still needs a way to determine which products people with similar preferences actually considered and eventually purchased.

Enter Criteo. It has spent years sitting between retailers, advertisers and shoppers, which gives it visibility into a huge amount of commerce behavior. Its systems can see products people browse, compare, add to their carts and eventually purchase across a large network of retailers.

Criteo says its platform touches roughly 720 million daily shoppers and more than $1 trillion of annual commerce activity. But Google already has an enormous shopping graph and Shopify has billions of products flowing through its ecosystem.

The more interesting asset to me is the outcome data. Knowing that a shoe costs $150 and comes in five colors is easy. Knowing that consumers who looked at three competing shoes repeatedly ended up buying one of them at a certain price is much harder to recreate across thousands of merchants.

Criteo is already trying to turn that data into infrastructure for agentic shopping. It launched an Agentic Commerce Recommendation Service that allows AI systems to use Criteo’s commerce intelligence to help rank products.

OpenAI also selected Criteo as its first advertising technology partner for ChatGPT Ads, and more than 2,000 brands were already accessing ChatGPT inventory through Criteo by June. That gives us some actual validation that the company has assets useful to the emerging agent ecosystem.

The current business is where I have much more hesitation. Q2 revenue declined 11% and adjusted EBITDA fell 18%. Management also reduced its full-year outlook after two large Retail Media customers cut the scope of their relationships. The agentic products are still early enough that they contribute very little to the financial picture today.

So the investment case depends on a transition. A big one. I would effectively be buying a struggling ad-tech business because I think some of the infrastructure and data it built for advertising become much more valuable as commerce shifts toward AI agents. If that transition takes years, or never produces meaningful revenue, the existing financial profile can continue weighing on the stock.

The competitive risk is also very real.

Google has its Shopping Graph.

Amazon has enormous amounts of purchase data.

Shopify sees transactions across millions of merchants.

OpenAI is taking product feeds directly and building its own advertising and commerce infrastructure.

Meta can do the same around Muse.

Criteo has useful data, but several of the largest technology companies in the world are building overlapping datasets.

So basically the version of this thesis that really interests me is Criteo becoming a neutral layer connecting advertisers and merchants across multiple AI agents. A merchant probably wants distribution across ChatGPT, Muse, Gemini and whatever other shopping interfaces emerge. Criteo already connects advertisers and retailers across a fragmented advertising market, so extending that role into agentic commerce would make sense.

That means I am watching for a few very specific things.

1 - more AI platforms integrating Criteo

2 - actual agentic revenue beginning to show up in the financials

3 - evidence that Criteo’s cross-retailer transaction data meaningfully improves product recommendations.

I also want to see whether advertisers increasingly use Criteo as the single interface for reaching multiple agent platforms. If that starts happening, CRTO could eventually look very different from the business investors are valuing today. If the major platforms build enough of the stack internally, the existing slow-growth ad-tech business remains the dominant story.

Sell side is still valuing Criteo almost entirely on the existing ad-tech business. After the latest earnings reset, most price targets sit around $20–$25, which only implies roughly 2.5–3x forward EBITDA depending on the estimate used. That is still a very low multiple and tells me analysts are mainly underwriting stabilization in the core business rather than assigning much value to agentic commerce.

That is what is interesting. I do not need Criteo to suddenly deserve a huge AI multiple. If Performance Media stabilizes, Retail Media gets back to growth, and the agentic products start becoming a real source of revenue, the market may eventually have to value those pieces separately instead of putting one depressed ad-tech multiple on the entire company. For now, I think that agentic upside is mostly optionality. Sell-side is still focused on whether the existing business can stop getting worse.

As of today, it goes on my research list.

What do you guys think? Still got plenty of digging to do for me. It’s new to me and an unfamiliar business model .

I think that OpenAI is going to push agentic shopping even harder after all this Muse stuff. If that happens it’s possible investors will start looking into second order beneficiaries like we are seeing with INOD with Muse for example. Curious if that could happen here.

Financial Disclaimer

This content is solely for informational and educational purposes and should be treated as general investment research rather than personalized investment, legal, tax, or financial advice. Any opinions, estimates, forecasts, valuation ranges, price targets, scenarios, or forward-looking statements reflect my views at the time of publication and may change as new information becomes available.

I may own, buy, sell, or otherwise have exposure to securities discussed in this publication, and those positions may change at any time without an update to this article. Readers should assume that potential conflicts of interest may exist whenever I discuss a public company or security.

All investing involves risk, including the potential loss of principal. Small-cap companies, emerging technologies, AI-related businesses, and companies undergoing strategic or financial transitions can carry especially high volatility and execution risk. Forward-looking scenarios depend on assumptions that may prove inaccurate, and actual financial results can differ materially from estimates.

Information is sourced from company filings, earnings materials, management commentary, industry sources, third-party research, and other publicly available information believed to be reliable. Accuracy, completeness, and timeliness cannot be guaranteed. Investors should conduct their own research and consider their individual financial situation, objectives, risk tolerance, and time horizon before making investment decisions.

Past performance and historical valuation relationships provide no guarantee of future results.