In an exclusive conversation with MartechAI’s Anupama Mitra, Todd Parsons, Chief Product Officer and President, Performance Media at Criteo, shared his views on the future of AI in commerce, the balance between regulation and innovation, and why performance-first measurement will define the next era of retail media.

Q. Recently, we’ve been hearing terms like empathy intelligence and emotional engineering in the context of AI. How do you see these concepts impacting commerce?

The concepts are fantastic, but the challenge is how you actually train a model to be empathic without fully understanding how the model functions. Right now, even the makers of these systems do not always know how they behave in autonomous settings. Certain behaviors can be hard-coded, but real autonomous empathy is not something we can control yet.

Research, especially around new models like GPT-5, continues to expose gaps in predictability. That makes safeguards critical. We need ways to monitor, test, and benchmark models. But it is still murky. Personally, I do not have a strong opinion as a researcher would, but I see intent as important. Intent drives innovation, even if the problem itself is almost impossible to solve today.

Q. Agentic AI is said to capture emotions and influence consumer behavior, sometimes even by reading facial expressions. How do you see companies using this data responsibly?

The first step is building a policy engine that governs how data can be used. That means row-by-row, column-by-column rules for what can be queried for training or retrieval. It is a return to fundamentals of database design.

From my experience at Acxiom, I learned privacy measures from highly regulated industries like banking and healthcare. Those principles are being reintroduced into AI. On top of that, you need abstraction layers, synthetic views of data that protect security but still allow functionality.

But there is a fine line. Too much density of data in vector space can feel like GDPR violations. This is why regulation matters. Europe is leading again with the AI Act, much as it did with GDPR. At Criteo, we start with what we cannot do, not what we can. It can feel limiting, but ultimately it ensures trust.

Q. You have often said AI is more than a feature, it is a foundation. What is your personal vision of how AI will redefine commerce in the next five years?

I believe AI will evolve into a channel in itself. It is aptly called a co-pilot. I do not think humans want agents making all decisions for them, and I do not believe zero-click commerce will replace retail.

Instead, AI will be another channel in the media mix. Think of ChatGPT as another publisher, a massive one, but not unlike Meta or TikTok. Where we will land is a channel context where agents perform autonomous tasks for certain trusted functions, while brands focus on discovery and attention.

This excites me because unlike traditional formats such as auctions, ad placements, or clearing costs, agentic AI involves completely new decisioning models. It is innovation from scratch, and that makes it a dynamic space for experimentation.

Q. How is Criteo approaching commerce personalization differently than big platforms like Meta?

Our focus is on holding performance constant across channels. Whether a user is on TikTok, Reddit, Zepto, or Instagram, the goal is to ensure measurement is consistent.

We do not want to debate measurement. We do the work, share logs with partners, and bring third-party validation so the results are trusted. That is value in itself.

Looking ahead, we are building a commerce DSP, a demand-side platform that simplifies decisions. Instead of complex dashboards, brands and agencies will be able to toggle channels easily. For example, exclude TikTok or prioritize Zepto due to a partnership. The philosophy is simple: performance-first, across every channel.

Q. What does Criteo’s API-powered commerce ecosystem look like in the next three to five years?

It is about automating how data powers shopper journeys. Our vision is to provide a data layer that supports not just advertising but also connected customer experiences, both onsite and offsite.

The goal is relevance without creepiness. That means enabling brands and retailers to decide, legally and ethically, how to use data for better CX. It is similar to how we approached retargeting and retail media, but broader. More use cases, more monetization opportunities, and more value for partners who need expertise.

Q. Today, AI is also being used to create campaigns in record time, sometimes within two days. What is your take?

It is exciting, and agencies like WPP are doing brilliant work here. But I think the question is always about the data quality behind these fast-turnaround campaigns.

AI makes building a DSP or ad server easier than ever. But the value lies in the quality of data and how it is distributed to partners. At Criteo, we stay focused on outcomes. For us, it always comes back to commerce lift. Did it lead to a sale?

That focus simplifies everything. Many companies think about media first. We think about sales outcomes. And that purity keeps us grounded.

Q. Finally, as formats evolve, do you think 30-second spots are still relevant?

I grew up with 30-second spots, but now I am retrained by the For You Page. For me, even 15 seconds can feel long.

So it is both generational and personal. Younger consumers may prefer shorter bursts. Others still value 30 seconds. What matters is quality and relevance. If it is meaningful, the format becomes less important.