Every product discovery experience built in the last decade assumes customers know what they want before they open an app. People have always shopped by feeling first — they just finally have a channel that speaks that language.  

Coffee, served by intuition  

On April 15, Starbucks become one of the first major restaurant brands to embed directly inside ChatGPT — currently available to US customers only. Tag @Starbucks in a chat, describe a mood or upload a photo, and get drink suggestions before heading to the Starbucks app to complete checkout.  

Paul Riedel, Starbucks' SVP of digital and loyalty, made the logic plain: "Customers aren't always starting with a menu. They're starting with a feeling."

That shift in where the journey begins is the whole story.

Product discovery is shifting  

Behind one chat prompt sits six years of infrastructure.  

Starbucks' proprietary Deep Brew AI platform has been running since 2019, processing over 100 million weekly transactions across 75 million loyalty profiles. Built on its own transaction data, it powers recommendation engines, forecasts inventory demand, and drives personalised offers across the Starbucks ecosystem.  

That foundation is what makes the ChatGPT integration credible. Starbucks can translate "I want something warm and cozy" into a confident suggestion because it already knows what their customers order, when, and why. The discovery happens in ChatGPT. The relationship is closed on Starbucks' terms.

Brands without that foundation are invisible the moment discovery moves upstream.

Walmart, Sephora, Target, and other major retailers are already building inside ChatGPT's app directory on the same logic. The pattern is still forming, but the direction is clear: customers are beginning to start their journey in AI tools rather than brand-owned apps.

Build the layer that lasts  

Any brand can plug into ChatGPT's app directory. What they cannot replicate in a sprint is the customer intelligence that makes the experience worth having — the purchase history, behavioural patterns, and preference signals that tell an AI not just what to suggest, but why that suggestion will land.

The AI enhancement is only as good as the foundation beneath it.

For enterprise teams, the more important question is whether the internal data layer is ready to support the next AI channel before committing to it. How well do you know your customers' behaviours across touchpoints? How structured and accessible is that knowledge to the teams building new experiences? How much of what your brand knows about its customers lives in systems an AI layer cannot reach?

New channels will keep opening. Conversational commerce, agent-based discovery, and tools that don't have names yet. Brands with deep, well-organised customer intelligence will move fast and well.  

Brands still assembling that foundation will watch from the side.