AIRBNB EXPERIENCES
Marketplace Strategy · AI Products · Supply Planning
Turning a curatorial point of view into an AI-assisted supply-planning system—helping Airbnb decide what Experiences to acquire next across markets.
IMPACT
Reduced supply priority development from 2–3 weeks to roughly 5–10 minutes for an initial portfolio—kickstarting ideation so that curators could focus their expertise on local nuance, emerging opportunity and the decisions that required taste.
SHAPING THE FUTURE OF THE MARKETPLACE
As Airbnb expanded Experiences, growth depended on more than increasing the volume of supply. Teams needed a disciplined way to determine which kinds of Experiences were missing or underrepresented in each market—and which opportunities were meaningful enough to pursue.
Offline Design helped shape those decisions by defining creative acquisition briefs describing a supply opportunity, the kind of Experience that could address it and the type of host or operator capable of bringing it to life. Creating a strong prospective supply portfolio required synthesizing several kinds of evidence: marketplace performance, audience insight, local context, business priorities and the editorial principles guiding how Experiences should show up in a city.
Done manually, that work could take two to three weeks for a single market. As planning expanded across more cities, the process itself became a constraint on the organization’s ability to move. That’s when I partnered with another curator to explore whether AI could accelerate the workflow without flattening the creative judgment and experiential expertise that made the briefs valuable.
ENCODING CURATORIAL PRACTICE
The opportunity was not to automate curation. Curators were closest to their markets and ultimately owned the work: they understood the local nuances, emerging scenes, cultural edges and less obvious opportunities that a generalized system could not reliably see on its own.
Instead, we focused on the part of the process that could be made repeatable. We translated the team’s shared knowledge into a structured model—drawing on marketplace evidence, audience research, trusted editorial inputs, city-specific principles and examples of strong supply—to give AI a credible starting point for scaffolding research and ideation.
The workflow could establish the common-sense baseline: what a market appeared to need, which opportunity territories were worth exploring and what an initial portfolio might look like. Curators then worked from that foundation—challenging assumptions, identifying what the system had missed and adding the taste, nuance and savoir-faire that came from actually knowing the market. AI provided the head start. Curators owned the point of view.
FROM PROMPT TO PRODUCT
The first AI-assisted workflow proved that this division of labor could dramatically accelerate the work, but it still required considerable prompt fluency and manual orchestration. Working with Design Operations, we then moved that complexity behind a simpler product UI that enabled curators to configure a market and its relevant context while the system handled the repetitive research, synthesis and first-pass ideation underneath.
The resulting recommendations were deliberately treated as starting material rather than answers. Curators reviewed, reshaped and approved the work before it entered the broader supply-planning process. Their role became more consequential, not less: less time rebuilding the baseline, more time interrogating what was obvious, finding what was missing and deciding which opportunities were culturally and commercially worth pursuing. The sophistication moved into the system so the curator could spend more time at the edge.
FROM WEEKS TO MINUTES
This workflow reduced the time required to develop an initial Supply Gap portfolio from roughly two to three weeks of manual work, to several hours with the early AI-assisted approach, and ultimately to around five to ten minutes through the productized workflow.
Speed was only useful if it protected the quality of the decisions. While the system automated the repeatable middle of the process, gathering established inputs, synthesizing evidence and producing a credible first pass, curators remained responsible for the part that could not be standardized—the emerging, locally specific and creatively consequential judgments that distinguished an adequate marketplace from an interesting one.
That changed where valuable expertise was spent. Instead of using curator time to repeatedly establish the baseline, the team could start from a stronger common foundation and devote more attention to taste, nuance, local intelligence and the opportunities the data had not yet learned to see.
What began as an experiment in acceleration became a model for a different relationship between AI and creative expertise: automate the common sense, then give experts more room for the things that only they can bring to the table .