Olly's Jennifer Peters on Healthy Skepticism & the 'Retail Media Magic' Ahead in the AI Space
The biggest takeaway for CPG brands navigating the exploding space of AI and large language models (LLMs)? Accept that nobody truly knows exactly how it will all pan out — and approach new tools with a healthy dose of skepticism.
"Remember that everybody's trying to sell you something," said Jennifer Peters, Olly director of direct-to-consumer marketing technology and digital compliance, during an episode of Retail Media Unplugged with Senior Editor Cyndi Loza.
Peters notes that while traffic driven by AI agents is growing exponentially, the most reliable strategy brands can execute right now is foundational: ensuring their product catalogs are fully digitized and data is clean.
"Anybody working in the AI space, we're making the best guesses we can make," Peters said. "So much of what we're doing, and what any brand can do, is really just good business. If you go through your website and make sure all your data is structured data, awesome. That's great no matter what happens ... in the space of LLMs."
Other topics discussed include:
- The "Retail Media Magic": When an AI agent recommends an omnichannel product to a shopper, how does it decide which retailer to send them to (e.g., Walmart, Target, Amazon, or DTC)? Peters predicts that this intersection — how LLMs display results and weight the importance of different channels — is exactly where the potential "retail media magic" could happen.
- The Problem with Product Titles: For years, the standard playbook involved tweaking product titles differently across Amazon, Walmart and Target to maximize channel-specific SEO. Now, LLMs, which read titles rather than UPCs, often can't recognize that these are the exact same item, forcing brands to step back and rethink their naming strategies.
- Agentic Commerce Realities: While the industry buzzes about a near-future where every consumer has a personal AI agent to do their shopping, Peters remains highly skeptical. She notes that until AI models can reliably parse highly nuanced product attributes without hallucinating or confusing items, human validation and traditional SEO remain completely essential.
