Stardrift AI
2025
Stardrift is AI travel search for frequent flyers, helping you plan and book complex itineraries. Users use it to find the flights that work for their schedules and preferences.
Figure 1. Stardrift Demo. The video demonstrates the features of the booking flow.
Detailed Experience Audit
I conducted an experience audit of the Stardrift website to understand the current user experience and identify opportunities for improvement. Here are the slides from the audit:
Figure 2. Slide describing the current agent process for booking complex trips.
Figure 3. Slide highlighting issues with the current booking page hierarchy.
Figure 4. Slide proposing improvements to the information hierarchy.
Figure 5. Slide calling out inconsistent formatting patterns in the interface.
Figure 6. Slide showing award flight search results and issues.
Figure 7. Slide reviewing booking flow hierarchy opportunities.
Figure 8. Slide comparing booking options and tradeoffs.
Figure 9. Slide suggesting more clearly clickable card components.
Figure 10. Slide summarizing general improvements to the booking flow.
Figure 11. Slide introducing a new booking experience concept.
Figure 12. Slide with additional new booking design explorations.
Summary of Experience Audit
Users cannot tell if the agent is actively working. Static text and icons make the system feel stalled. Agent status copy should update in real time to explain what the agent is doing.
Numbers lack clear labels such as seats. Additional fees are inconsistent between screens, creating distrust. List based results make comparison slow and mentally taxing.
Booking buttons look like headers instead of actions. Primary actions are not visually prioritized.
Multiple bullet styles are used. Date formatting varies and feels non standard. Reasoning text looks too similar to output text and blends together.
Tables would make flight comparisons easier. Consistent formatting would improve scannability and perceived quality.
The line and dot motif works but should be simplified and used more deliberately. Reasoning content should be visually separated, for example using block style formatting.
Reflections
I want to thank Leila Clark for giving me the opportunity to work on Stardrift in its early stages. It was a great learning experience and I'm grateful for the opportunity to work on such an impactful product.