Optivoy
An AI travel-planning copilot that helps people create, understand, adjust, and manage a trip — all in one place.
Most AI planners stop at generating an itinerary. The harder, more interesting problem is what comes after: helping someone understand the plan, weigh its trade-offs, and change it without starting over. That framed the central design question:
How might an AI travel planner recommend a trip without taking control away from the traveler?
Home opens with the state of your travel life, not a search box: your next trip and how ready it is, one action to start something new, and a running tally of what the AI has already saved you. Its work always shows up as suggestions you can review — never silent changes.
Planning a trip means holding too many things together.
I started by looking at how people actually plan a trip today — and where the effort really goes. Four problems kept showing up.
Designing around one realistic trip.
To keep the work grounded, I designed against a single, concrete trip instead of an abstract audience. This is a scenario I used to guide the concept — not researched user data.
Alex has a real budget, a few free evenings to plan, and a partner with slightly different interests. They want recommendations they can look over and adjust — not a plan they're expected to accept on faith.
The first question wasn't what the UI should look like — it was what the product should actually do.
The first version could have stopped at "enter your details and generate an itinerary." But that just makes another one-time AI generator. I wanted the product to keep working after the plan appears, so I mapped a broader flow before touching any screens.
The last two steps are where most tools stop — and where Optivoy's real value lives.
Before polishing the UI, I mapped the core flow.
I sketched the whole path in low fidelity first — from opening the app to managing the trip after it's built — to make sure the structure held before any visual design.
The wireframes helped me see that the product couldn't end at generation. Creating the plan was only the midpoint.
From an itinerary generator to an ongoing trip copilot.
Three shifts happened between the low-fidelity flow and the final design. Each one moved the product further from "generate and done."
Instead of one big form, the user answers one kind of question at a time: trip basics, then preferences, then budget, then review.
Each recommendation now shows its reason, its benefit, its trade-off, and a clear Apply action — instead of a plan you either take or regenerate.
The itinerary, budget, preparation, and preferences stay connected after generation — one trip you keep editing, not a document you export.
Four short steps, in the order people can answer them.
Facts first, then taste, then money, then a review before the AI builds anything. Each step is one screen with one job.
A recommendation should explain itself before asking for trust.
I wanted the AI to make its case, not just change the trip. Every suggestion carries five parts, and applying one is always a deliberate, reversible choice — the AI proposes, the traveler decides.
Apply is deliberate and reversible. The AI proposes; the traveler decides, one suggestion at a time, and can undo any change afterward. The goal isn't for people to accept everything — it's for them to accept the suggestions that are actually right for their trip.
Adjusting the plan should feel like editing — not starting over.
Rather than a "regenerate" button, the user adjusts named priorities — budget, pace, schedule density, travel radius, local vs popular — and sees the effect before committing to it.
The trip stays useful after the AI finishes.
One trip carries four tabs under a persistent header — so planning and managing never feel like separate apps.
What happens when the recommendation or plan is wrong?
An AI product only earns trust if it handles the moments it gets things wrong. These are the states I designed for.
What I would test before treating the idea as successful.
This is a concept built on product analysis and assumptions, not completed research — so here's the plan, not the outcome. I'd sit with travelers planning a real personal trip and watch for five things.
What designing Optivoy taught me.
AI works better as decision support than as an authority. The AI proposes, explains, and admits trade-offs; the person decides. Every screen follows from that.
Personalization and control belong together. A plan feels personal when your priorities visibly shaped it — and you can re-shape it whenever you want.
Optimization has to be visible. Live previews, signed changes, and before-vs-after comparisons turn invisible work into something people can trust.
The experience can't end at generation. Preparation and everyday trip management are the core of the product, not an afterthought.