By Shannon Millard, VP Platform, Epsilon
AI is changing how we shop and search, but not always how marketers think. Agentic commerce, powered by conversational AI, promises a future where consumers simply ask for what they need and the system does the rest. It sounds frictionless, but the reality is more complex.
72% of people already use AI tools for search, yet those interactions generate less than 1% of traffic. These are fleeting moments of curiosity, not enduring connections. Prompts reveal intent in that instant, but they strip away the continuity that turns one-off questions into long-term relationships.
Real discovery has always depended on more than a single line of input. It’s built on the context of the meals people plan, the brands they trust, the habits they repeat. Those signals don’t live in chatbots, they live in purchase histories, loyalty programmes and the many touchpoints that record how people actually behave.
The limits of prompt-based discovery
AI prompts are framed by what consumers already know or imagine. A shopper might ask, “Show me brown boots under £100,” but that query doesn’t reflect that they’ve been browsing premium brands for weeks or regularly buy sustainable labels. Agentic interactions collapse all of that nuance into an all-to-often price-driven exchange.
That makes discovery narrower, not richer. Instead of expanding horizons, AI search reinforces assumptions by showing what’s most reviewed, cheapest or closest, not what’s genuinely most relevant. The risk for brands is obvious: over-invest in these systems and you train consumers to think in shortcuts.
Where discovery really happens
The more powerful opportunities lie in environments that understand attention and context. Channels like connected TV, audio and streaming are simultaneously places to be entertained and where curiosity and inspiration thrive. These are moments when people are open to ideas like retro fashion, new home style elements, recipes on screen, a travel documentary that sparks a booking.
Advertising here should feel part of that experience, but too often it doesn’t. Many campaigns are still shaped by legacy publisher deals or blunt audience segments, so the creative feels disconnected from what people actually care about. With identity-based targeting and first-party data, those same environments become fertile ground for discovery, connecting what a person is watching or listening to with what they’ve recently browsed, bought or searched.
Retailers hold the missing link
Retailers are uniquely positioned to bring that connection back. Loyalty and transactional data provide a real-world map of what people buy together, how often, and in which categories. Pairing that intelligence with CTV or audio activation allows brands to engage shoppers in genuinely relevant ways.
A wine brand could reach customers who recently bought steak, with creative served during a cooking show. A technology brand could reconnect with shoppers who browsed headphones through dynamic podcast ads that demonstrate noise-cancelling sound during rush hour. When that same data powers measurement, brands can see whether the ad drove an in-store sale or boosted repeat purchase, turning inspiration into outcomes.
Why AI still needs people
Predictive AI becomes far more powerful when it’s guided by identity and real-world signals. It can learn from loyalty and purchase data to anticipate needs, adjust creative and decide when and where to show it. In that sense, AI’s future depends less on responding to prompts and more on interpreting people.
Retailers that invest in this kind of person-first data strategy will be able to compete not just with each other, but with the agentic platforms forming around them. They’ll retain control of the customer relationship and keep brand discovery flowing through channels where they can see, measure and shape engagement, rather than handing it over to a third-party interface.
The next wave of discovery
As AI becomes more capable, it will take on a larger role in search, service and commerce. But brands and retailers shouldn’t over-index on it. The better route forward is balance: embrace agentic tools where they add efficiency, but keep investing in the contextual, identity-driven environments that reflect how people really discover and decide.
The truth is, discovery has never been about one click or one conversation. It’s about recognising the person behind the prompt and using data, creativity and technology together to meet them where intent and inspiration collide.
