As AI reshapes product discovery, commerce and media buying, Criteo’s International Managing Director, Marc Fischli believes the enduring competitive advantage will not necessarily be the newest technology, but access to the right data. He also sees traditional marketing silos beginning to blur as campaigns become increasingly cross-channel and AI-driven.
The pace of change in marketing has become so rapid that conversations that felt futuristic just a year ago are already becoming part of everyday business.
Marc Fischli, Executive Managing Director, International Markets at Criteo, offered a striking example from Cannes. After participating in a panel with travel industry clients, Fischli and the other panellists discussed how much of what they had just spoken about would even have been part of the conversation a year earlier.
Their estimate was roughly 80%.
“That’s the speed of change,” he said.
In an exclusive virtual interview with Brij Pahwa, Editorial Lead, exchange4media Group and Editorial Head, MartechAI.com, Fischli spoke about the rise of agentic AI, changing consumer discovery behaviour, the growing importance of commerce data, the limits of autonomous agents, India’s digital ecosystem and why the structure of marketing organisations themselves could look very different within the next three years.
At the centre of this shift is the transition from generative AI experimentation towards real applications.
“I think 2026 is definitely the year where we see the applications come through,” Fischli said.
Criteo, he explained, is investing specifically in generative and agentic AI and integrating those capabilities into its products. One area showcased by the company is the ability to make capabilities available through Model Context Protocol, or MCP, allowing them to connect directly with client platforms and interfaces, including commercially available large language models.
According to Fischli, this could allow marketers to launch and run campaigns through these emerging interfaces.
Discovery is changing faster than transactions
For all the excitement around autonomous commerce, Fischli draws an important distinction between discovery and transaction.
AI systems and LLMs are already changing the way consumers find products, compare alternatives and narrow down choices. However, consumers are not necessarily handing over the entire purchasing journey to an AI agent.
“We see the development of AI-driven commerce as a new ability for marketers to become more efficient and an ability for consumers to have easier access to look for, search, find, discover and ultimately purchase,” he said.
For now, Fischli largely sees AI as an additional channel.
Discovery is moving rapidly into LLMs and AI-driven interfaces, while many transactions continue to take place with the original retailer, e-tailer or travel company.
There is a commercial reason for that. The transaction creates valuable information and becomes the beginning of the next data cycle. Retailers and travel companies, therefore, have little incentive to simply surrender that relationship.
The same distinction applies to fully autonomous agent-to-agent commerce.
While the technology required for AI agents to negotiate and execute more activities is increasingly available, Fischli said Criteo is not yet seeing significant movement towards a world where agents independently handle the entire commercial process.
He also argued that many companies developing large language models are currently focused on improving their models, achieving scale and remaining competitive in the AI race rather than immediately trying to take control of every transaction.
From SEO to optimisation for machines
The evolution of discovery creates another challenge for marketers.
For years, brands learnt how to optimise for Google. More recently, they have begun thinking about visibility within AI-generated answers. Agentic discovery now introduces another layer in which the audience consuming product information may not always be human.
Fischli believes the fundamentals remain surprisingly familiar.
“At the end of the day, if we take a step back from the three-letter codes and what we in the industry like to talk about, it’s about how do you, in a positive sense, influence shoppers to go towards your product rather than a competitor’s product,” he said.
The difference is that brands may increasingly need to provide information that machines can understand and evaluate.
Product information becomes particularly important. Fischli pointed out that while a retailer’s homepage and search function may once have been the primary starting points, consumers can now arrive directly on product pages after discovering products elsewhere.
That makes the quality, language and findability of individual product pages considerably more important.
AI also needs information beyond basic specifications.
Ratings, reviews and platforms containing consumer opinion can provide signals around product quality and the more emotional aspects of purchasing decisions. For travel companies and online travel platforms, much of that information may already exist within their own ecosystems. The challenge is ensuring that LLMs and agents can access it appropriately.
The AI race may ultimately be a data race
Fischli repeatedly returned to one issue throughout the conversation: data.
Criteo’s commerce intelligence infrastructure includes its shopper graph alongside product catalogue information at the SKU level, allowing it to understand interactions as well as product-level information across its ecosystem.
But Fischli does not believe any single company can necessarily provide every type of data a marketer needs.
“The industry has become really complicated and there’s no single company who can do all the things themselves,” he said.
That makes partnerships increasingly important.
He described data through three broad dimensions: breadth, depth and quality.
A platform operating across a large portion of the commerce ecosystem may provide significant breadth. A retailer, however, could possess much greater depth through loyalty and customer data. A consumer goods company may have detailed knowledge about its own brands, market share and consumer perception.
Depending on the objective, one dataset may be sufficient. In many other cases, marketers will increasingly need to combine multiple datasets.
That is why Fischli believes the ability to build the right partnerships could become a major marketing capability.
AI has transformed discovery, but trust remains a barrier
Asked what Criteo’s view of global commerce data suggests about changing consumer behaviour, Fischli identified discovery as the biggest shift.
Consumers can increasingly use natural language and even voice to describe precisely what they want, allowing AI systems to narrow down options much more quickly than traditional search processes.
Travel is one obvious example.
Consumers are already using LLMs heavily to build itineraries and research trips. But allowing an AI agent to complete the purchase is another matter.
The barrier is trust.
Consumers still have questions about whether they should allow an AI system to handle credit card details or other financial information, whether it will select the correct dates and whether it will execute exactly what they intended.
Fischli said Criteo’s data indicates that 96% of people using LLMs also use another channel during their purchase cycle.
That suggests AI discovery is expanding rapidly without necessarily replacing the rest of the consumer journey.
Agentic AI needs boundaries
Greater autonomy also creates new ethical questions.
Asked about situations in which autonomous agents could go beyond the user’s intended objective while trying to complete a task, Fischli acknowledged the need for guardrails.
“Like many new technologies, [it] will pose some ethical questions,” he said.
He argued for checks and balances as well as independent bodies capable of examining how these systems operate.
For Fischli, one guiding principle should remain relatively simple: technology should ultimately serve the consumer.
“As long as we’re true to ourselves as an industry about really serving the customer, the consumer, the shopper, we’re going to be broadly right,” he said.
India’s advantage is speed, but fragmentation remains
Having worked across international markets, Fischli does not view the evolution of AI and commerce as a straightforward race between countries.
India, however, has some distinctive advantages.
He described it as an “entrepreneurial” and “very fast-moving environment”, qualities that become increasingly valuable when adaptability itself is becoming a competitive advantage.
India can also take something relatively small and bring it to scale quickly.
The challenge is fragmentation across different parts of the value chain. Maintaining scale while establishing common standards and easier ways of accessing technologies, services and products remains an ongoing challenge.
Despite that, Fischli described India as an “exciting market” because of its underlying adaptability and entrepreneurial culture.
The marketing organisation itself could be next
Perhaps the most significant disruption may happen inside companies.
Asked what could appear outdated if the same conversation were held three years from now, Fischli was reluctant to predict individual technologies in an industry moving so quickly.
Instead, he focused on organisational structure.
Today, many companies operate separate performance marketing, search, social and shopper marketing teams. But if campaigns are increasingly planned and bought across channels, those divisions may become increasingly difficult to maintain.
“At the very least, these groups of people [need to] start to talk and work much more closely together,” he said.
Looking three years ahead, Fischli went further.
“I’ll be surprised if these teams would still exist in the current form.”
The tools, interfaces and channels may change dramatically. But in Fischli’s view, one fundamental is less likely to disappear: access to the right data, with the right quality, breadth and depth, for the objective a marketer is trying to achieve.
In an industry where 80% of today’s conversation can seemingly emerge within a year, that may be one of the few things marketers can reasonably expect to remain constant.
Related Articles
