India is emerging as a testing ground for agentic artificial intelligence (AI) and the next wave of commerce innovation, according to Todd Parsons, Chief Product Officer and President, Performance Media at Criteo. Speaking to media, Parsons described India as not only one of Criteo’s fastest-growing markets but also one of the world’s most dynamic ecosystems for AI adoption. “India represents a confluence of quick commerce, a youthful digital population, and a willingness to experiment with AI systems that we’re not seeing elsewhere,” said Parsons. “It’s incredibly exciting because these behaviors drive how we think about product development globally.”
Also read: Criteo and Zepto Partner to Drive the Next Wave of Quick Commerce in India
Parsons highlighted a striking data point: India is far ahead of the United States in AI adoption. “About 13.5% of global monthly active users of ChatGPT are here in India, compared to just 8.9% in the U.S.,” he noted. “That’s 500 basis points higher, and it speaks volumes about India’s digital-first behavior.” He also cited projections that India’s digital commerce market could exceed $1 trillion by 2030. “For someone leading a P&L and building product strategy, those numbers are intoxicating,” Parsons said. “It’s a signal that the time to invest here is now.”
The Rise of Agentic AI and Why Trust is the Key
Parsons sees agentic AI and AI-driven agents performing tasks end-to-end as the next major shift after search, social, and mobile. But the technology is not fully mature yet. “In just two years, ChatGPT went from a nerdy experiment to serving 10% of the global population. I’ve never seen an adoption curve like that. In fact, not even with YouTube, Facebook, or WhatsApp,” Parsons said. “But while curiosity is high, trust is lagging. People are using AI to discover products, but they hesitate when it comes to completing transactions.”
According to Criteo’s own research, 85% of users are using AI tools for shopping, and 71% are doing so weekly, yet conversion rates remain low. “The funnel collapses at the point of purchase,” Parsons explained. “Users are asking AI to recommend products, but they still want to verify before spending their money. That’s a trust gap we need to close.”
Parsons outlined four critical conditions for agentic AI to gain consumer trust:
- High-quality data inputs to eliminate hallucinations.
- Reliable subroutines connecting personal agents with transaction agents.
- Broader context windows so AI can truly personalize recommendations.
- Robust policy enforcement to protect data and prevent misuse.
“It takes a village to make systemic change happen,” Parsons said. “Technical complexity, ecosystem coordination, and governance all need to come together.”
Personal AI Assistant for Every Shopper
In the near future, many consumers will rely on personal AI shopping agents that make decisions and complete tasks on their behalf, feels Parsons. There are four ways in which personal shopper Al agents may work:
- Draws on extensive range of data from memory.
- Finds an optimal solution.
- Connects to task-specific agents.
- Takes actions as authorized.
Parsons added that consumers can expect to see three major use cases driving this agentic future:
- Prompt-based product recommendations tailored to individual preferences.
- Smarter real-time audience activation for marketers.
- Agent-enabled full-funnel campaigns that automate everything from awareness to purchase.
Will AI Agents Be Able to Influence Shopping Decisions?
Discussions around AI often highlight risks such as model poisoning, particularly given the autonomous nature of agentic AI systems. This raises an important question: could these personal AI shopping agents be leveraged by brands to unduly influence or manipulate consumer purchasing decisions? Manipulation is already possible today, if permitted, technology can indeed be used to persuade or sell. However, the question highlights a broader concern: the crossing of ethical boundaries, opines Parsons.
“Agent-to-agent engineering acknowledges these risks but is still at an early stage and has not yet developed adequate safeguards. I firmly believe that personal agents will eventually become a reality, although not everyone shares this perspective. What remains underdeveloped are the safeguards and policy frameworks required to govern how these agents behave,” he highlights.
Parsons further added: “We have been examining this issue for quite some time, including during our work with Scale AI prior to its acquisition by Meta. This technology is still nascent. We maintain a dedicated policy and ethics team that rigorously evaluates data use cases through an ethical lens to ensure that fundamental principles are not compromised. While it is impossible to achieve absolute protection or perfection, we continue to invest in this area precisely because the technology has not yet matured to the point of offering comprehensive safeguards.”
Criteo’s Strategy: Data Depth, Not Data Scraping
One of Parsons’ central messages was Criteo’s focus on “permission data”, which is information provided directly by partners rather than scraped from the open web. “Some would say crawled data is being stolen. Retailers, publishers, and brands aren’t happy about it, but they need to be discovered in AI-powered search. At Criteo, we take a different path,” Parsons emphasized.
“Our number one priority is working with holdout data: stock levels, dynamic pricing, product catalogs, and data that models cannot just crawl. This makes AI recommendations accurate, trustworthy, and valuable to both consumers and brands.” Criteo uses Model Context Protocol (MCP) servers to integrate this data directly into AI workflows, enabling richer inference and reducing the need for complex API connections. “This is the infrastructure that will make agentic commerce possible,” Parsons said.
Experimenting with AI Advertising With India as a Living Lab
Criteo is already experimenting with AI-powered advertising campaigns, particularly in India. “Discovery advertising is evolving,” Parsons explained. “We can now infer intent without traditional behavioral signals. Agentic campaigns allow us to build creative dynamically, and not just static display ads, but video and multi-format campaigns.”
Parsons hinted at the future of “Agentic Engine Optimization” (AEO), a next-generation equivalent of SEO. “Brands will need to optimize for generative AI and agentic systems, not just search engines. That means new ways of labeling and annotating product data so it’s discoverable by AI assistants.”
Publishers, too, are feeling the shift. “I speak to global publisher executives every week. They’re worried about traffic siphoning, but also see opportunity,” Parsons said. “If content is formatted in a way that AI agents understand and can cite, publishers remain central to the buyer journey.”
The Human Factor Remains
Despite the promise of AI agents, Parsons believes human decision-making will remain a key part of commerce. “I don’t care what the agent tells me. Irrespective of that, I’ll still go on Reddit to read reviews. People will still read trusted publications,” he said. “Agentic AI will enhance discovery, but it won’t replace human curiosity and validation.”
Parsons predicts that fully trusted, end-to-end agentic transactions are “a few years away,” but the groundwork is being laid today. “Memory windows are expanding, data security is improving, and retailers are preparing. When these pieces align, agentic commerce will take off.”
For Criteo, India is more than a growth market. He reiterated that India was in fact a proving ground. “The pace of experimentation here is unmatched,” Parsons concluded. “India is not just participating in the AI revolution. In many ways, it’s leading it.”
