Travel & group experiences
Trek Travel AI agent system
A marketing and sales agent that turns a short request into a finished newsletter, itinerary, or handout.
Saved agent-created outputs. Review actions stay in your browser.
Newsletters, flyers, and itineraries stay together in a searchable library, ready to reopen and reuse.
Explore three agent-created assets
Newsletter
Christmas Lights Await
The actual agent-created newsletter: an opening letter, upcoming trips, generated posters, and supporting stories for senior groups.
Saved from the product library: Christmas Lights Await: Branson & Biltmore.Open the newsletter ↗
Saved agent-created output. Original artwork and layout; live contact links and tracking are disabled in this portfolio copy.
Why I built it
Less time making materials. More time selling trips.
The owner was writing copy, finding images, formatting assets, and coordinating delivery by hand. He needed finished marketing materials, with a way to revise and approve them.
My role
Client discovery, scoping, architecture, implementation, testing, evaluation, and delivery.
How I built it
- 01
Give the agents business context
I built TypeScript agents with the company’s trip information, audience, voice, and tools.
- 02
Turn requests into finished assets
The backend creates copy and visuals, renders the asset, and saves the work in a shared library.
- 03
Keep the owner in control
Revisions preserve context. Approval rules control delivery, and saved work survives interrupted sessions.
Testing and technical decisions
I test content quality, rendering, saved-work recovery, and approval gates. Sentry and Langfuse help trace failures and model cost. The application owns the business state; changing model providers still requires adapter changes and validation.
The value
A few prompts to a finished asset.
The owner can create marketing materials with a few prompts, review the finished asset, and get back to selling trips.
Client-reported: approximately 30 minutes saved per marketing asset.