Intro

Celeste AI was built to reduce the friction between marketers and their data. Instead of navigating multiple views, filters, and reports, users can explore performance through a conversational interface that adapts to their workflow and context. For marketing agencies in particular, this dramatically shortens the time spent on recurring reporting, allowing teams to generate analyses, summaries and client-ready insights in a fraction of the usual time.

As the design owner, I defined the vision, shaped the UX flows and led Celeste’s evolution from the simple chat widget inside one feature to the cross-platform conversational assistant capable of remembering preferences, visualizing data, handling files and automating tasks.

Challenge

Marketing teams spend a significant amount of time navigating dense data views, creating reports, manually comparing timeframes and searching for signals hidden in performance noise. While conversational AI promised relief, a generic chatbot would not be enough. The system had to understand marketing language, respect custom naming conventions, stay aware of context, handle partial or missing data, and present outputs in formats that marketers could actually use.

At the same time, the AI experience needed to live inside the product without disrupting existing workflows. Celeste had to feel like a natural extension of the platform — not a separate tool, not a gimmick, and certainly not something that required learning an entirely new way of working.