Criteo’s agentic AI revolutionising retail media with smarter, faster automation

Criteo advances its Commerce Media Platform with a shift towards ‘agentic’ AI that promises to automate decision-making, personalise experiences, and optimise retail campaigns in real time, raising both opportunities and concerns in the evolving digital advertising landscape.

The modern path to purchase is messy, fast-moving and resolutely multichannel. Consumers flit between retailer storefronts, social feeds, publisher pages and apps, leaving behind a scat...

ter of cookies, CRM identifiers and device signals that rarely line up. According to the original report, Criteo argues that the answer to that fragmentation is not more manual stitching but a commerce‑native layer of artificial intelligence that can translate fractured signals into actionable outcomes.

Criteo positions its Commerce Media Platform as that layer. The company says its models are trained on near two decades of commerce data , including what it describes as 4.5 billion SKUs, some 3,700 product categories, 17,000 advertisers and roughly 720 million daily active users , and that this SKU‑level depth allows its systems to infer intent and forecast buying outcomes with greater granularity than demographic proxies. Industry recognition has followed: Criteo’s DeepKNN embedding approach won an industry award in 2024, and an IDC MarketScape assessment named the company a Leader among worldwide retail media network service providers in October 2024. Those distinctions underline the commercial momentum behind the argument that retail media is a natural proving ground for more advanced AI workflows.

What Criteo describes as a shift from predictive to “agentic” AI is central to its roadmap. Where earlier systems advised , offering bids, recommendations and creative variants , the agentic layer is framed as capable of deciding and acting within marketer‑defined guardrails. The report outlines practical examples: generating a full campaign from a natural‑language brief, building and continuously updating audience segments from intent signals, and delivering plain‑English explanations of campaign performance tied to specific products and categories. Criteo stresses that this agentic layer coordinates rather than replaces existing automation, surfacing long‑standing capabilities such as instant creative generation and real‑time optimisation.

The company’s recent commercial moves illustrate how that ambition is being operationalised across the market. In June 2025 Criteo announced a global partnership with dentsu to integrate Criteo’s platform with dentsu’s services, promising AI‑driven audiences, buying tools and measurement to connect retailers and publishers with incremental brand spend. The same month Criteo introduced Auction‑Based Display technology to bring programmatic, biddable trading to retail media environments , a response to advertiser demand for flexibility and one already adopted by retailers including Costco and Shipt, the company said. Those products aim to marry retail media’s product‑level signal richness with the flexibility advertisers expect from open‑web programmatic channels.

Criteo’s platform is not presented in isolation. Its Commerce Max offering, launched in 2022, already targeted enterprise marketers with a self‑service DSP that spans on‑site sponsored placements and off‑site programmatic extension, touting access to billions of SKUs and hundreds of retailer relationships. More recently, the company has extended its ecosystem through partnerships and integrations: a 2024 deal to power David Jones’ retail media, the June 2025 dentsu collaboration, and in November 2025 an integration with Xnurta that lets advertisers manage campaigns across Criteo’s 225 retail networks from a single omnichannel dashboard. Collectively, these moves reflect an attempt to reduce the operational friction agencies and brands face when buying across many retail networks.

Criteo frames these technical and commercial advances within a responsibility narrative. The company says its AI Lab of more than 120 specialists runs over 100,000 A/B tests a year and refreshes models as frequently as every six hours to align output with marketer expectations. It points to collaboration with France’s INRIA through a FAIRPLAY research group on privacy‑preserving AI and to an internal Regulatory AI Group that monitors emerging rules; Criteo also hosted a companywide “Trustworthy AI” symposium in 2025. Such safeguards are presented as essential, because the move from advisory systems to agentic autonomy raises familiar concerns about transparency, data protection and “black box” decisioning.

There are reasons for marketers to approach the agentic pitch with measured interest rather than unquestioning enthusiasm. Agentic systems can streamline execution and surface insights faster, but they also concentrate decision authority in algorithms and require robust governance, clear performance measurement and careful handling of first‑ and third‑party data. Industry data and vendor roadmaps show rapid productisation of AI capabilities in retail media, but that also heightens the need for independent measurement and interoperable controls so that spend truly follows verified shopper intent rather than optimised proxies that drift from commercial reality.

For brands, retailers and agencies the practical stakes are large. If agentic systems can reliably create, target and adapt campaigns in real time while preserving privacy and explainability, they promise more relevant experiences, uncovering incremental demand and shortening the path from impression to sale. If they fall short, the danger is wasted spend and opaque attribution.

The marketplace is already moving in both directions. Criteo’s recent product launches and partnerships signal a push to make retail media pliable, programmatic and AI‑driven; analysts and clients will be watching how those capabilities perform at scale. As retail environments demand rapid optimisation across thousands of SKUs, the question is not whether AI will play a role , it already does , but how agency, transparency and measurement will be maintained as decisioning shifts from humans to increasingly autonomous systems.

Source: Noah Wire Services