Agent field notes · Travel
Same trip.
Different copilots.
Three approaches to helping you get away. A little research for your next build.
Lola
Lola describes an assistant for experiences: ask for a trip, an event, a stay, or a restaurant with particular requirements. Its FAQ says it compares options from several sources, explains its suggestions, and shows live prices. It currently runs in the browser; the FAQ says a mobile app is planned for 2027.
- Where it starts
- A request and your preferences.
- What we’d test
- Can it plan a concert weekend while respecting a hotel budget and neighborhood?
- A builder’s question
- Could a narrow experience—like a concert trip—be a clearer entry point than “plan anything”?
Mindtrip
Mindtrip presents a personal travel assistant that learns preferences, helps choose flights, stays and restaurants, and keeps trip details together. Its site also describes agents that watch for changes such as price drops and availability. Those are product claims; we haven’t measured alert accuracy or booking success.
- Where it starts
- An ongoing trip and the details that matter to you.
- What we’d test
- Does changing a flight budget update the plan consistently? Can we trace an alert to its source?
- A builder’s question
- What useful job remains after the itinerary is generated?
Layla
Layla’s current site describes a combination of AI planning and human travel advisors. The AI builds around dates, budget and style; an advisor then checks details and handles booking. That human involvement matters when comparing it with self-serve assistants: we should record advisor time separately from automated work.
- Where it starts
- A trip brief, followed by an advisor.
- What we’d test
- Which constraints survive the handoff, and what does the traveler have to repeat?
- A builder’s question
- Where does a thoughtful human handoff improve trust more than another automated step?
One brief. A fairer comparison.
Proposed test brief—not a completed test:
Plan three nights in Chicago for two adults, June 11–14, 2027. Hotel budget: $250 per night including taxes. Stay within a 20-minute walk of Millennium Park. Include one vegetarian dinner and show sources for prices and opening hours.
Use the same brief and follow-up change for every product. Stop before a purchase, reservation, or message to an advisor. Mark unavailable capabilities as “not tested,” never as a failed task.
- Constraints: Did the dates, full hotel budget, walking distance, and dietary request survive?
- Evidence: Do links support prices and opening hours at the time checked?
- Revisions: Lower the hotel budget to $200. Does the whole plan stay consistent?
- Effort: Record elapsed time, paid tier, retries, and human involvement separately.
Repeat each available task in fresh sessions before drawing conclusions. Publish prompts, dated outputs, access limitations, and verification notes with the results. Store keyword positions measure discovery; these tests would measure task performance.
Independent editorial research using the official sources linked above. Product capabilities change. No hands-on scores or customer usage estimates are implied.