Sanchez presents Adobe's experiment with websites that adapt to an individual visitor while they browse. Instead of generating an uncontrolled replacement for the whole site, the system personalizes selected content blocksAI personalization adapts information, recommendations, interfaces, or agent behavior to an individual or group using permitted context and observed needs. and grounds every variation in the organization's existing site corpusGrounded generation produces an answer from supplied, retrievable, and attributable evidence rather than relying only on a model's learned patterns. so brand rules and source material remain in control.
Signals such as visited pages, dwell time and a visitor's current query are used to infer intentUser intent inference estimates the goal behind a person's words or actions when the request does not state every relevant detail.. That intent can reshape search results, recommendations and a dedicated page for one person. In the demonstration, a query about making coffee while camping produces copy and product recommendations tailored to that exact need without requiring marketers to author thousands of predefined variants.
Low latency is essentialLow-latency AI inference produces a model result quickly enough for an interactive or time-sensitive application. because a personalized page still has to feel like a normal website. The team continuously evaluates prompts across models and providers for both accuracy and speed, with requirements varying by site. Sanchez reports generation near one second with a fast inference setup, making real-time assembly practical rather than a background process that arrives too late.
Marketers define goals in natural language and use analytics to refine the personalization loop, while the system can also create a working demonstration for a new site in under an hour. Sanchez argues that this points beyond conventional audience segments toward an audience of oneOne-to-one personalization adapts an experience for one individual and their current context instead of assigning them only to a predefined audience segment., where an assistant can compose a relevant experience for the current request, device and context. Closing thanks and applause are omitted.
Watch on YouTube
