3. Cicero — The AI travel companion

Every city hides the stories only a local knows. Cicero finds the ones made for you, tells them the moment you're in front of them, and turns your whole trip into a memory worth keeping.

Comment

Inspiration

You arrive in a city you've dreamed about. Within an hour, you're walking past the obvious — the same five landmarks everyone photographs — while the soul of the place stays invisible. The courtyard around the corner. The story behind that fresco. The piazza that comes alive at sunset.

We built Cicero because we felt this ourselves. We traveled, but we didn't really see the places we visited. We came home with camera rolls full of blurry photos and no strong memory of what any of it actually meant. Ask ChatGPT or TripAdvisor, and you get the postcard. Never the city.

The question we kept asking: what if you had a friend who actually knew the place? Not a tour guide reading a script, but someone who knew you — where you're from, what moves you — and leaned in at exactly the right moment to tell you what you were looking at?

That's Cicero.

## How We Built It

Cicero is a four-person, vertically integrated team, each paired with an AI copilot, shipping on 28-day sprints.

Stack:

  • Frontend: Flutter (BLoC architecture)
  • Backend: Firebase / Firestore
  • Maps: Mapbox for live positioning and routing
  • Intelligence layer: LLM APIs (Claude, GPT-4) for story generation and real-time personalization by visitor origin, interests, and context
  • Audio: ElevenLabs for every audio guide, in every language — the entire sound of Cicero runs on ElevenLabs
  • Contextual engine: An on-device "next best place" system that factors location, time of day, weather, energy, and what's actually open — proactively, without the user asking
  • Content architecture: A modular "lego block" audio system that lets us A/B test stories paragraph by paragraph across 100–300 culturally relevant places per city

Every piece of audio a user hears — across all languages, all cities, personalized by country of origin — is generated and delivered through ElevenLabs. It is the voice of Cicero.

## Challenges

Two things were genuinely hard.

1. The contextual engine. Getting the right story to the right person at the right moment — proactively, without the user asking — required building a system that reasons across location, time, weather, pace, and what's physically accessible in real time. The challenge wasn't just technical. It was deciding when not to speak. A companion that talks too much is worse than no companion at all.

2. Owning the long tail. The top five landmarks are easy. The next 295 places in a city are not. We had to build an AI content pipeline that covers the full cultural surface of a city, personalized even by the visitor's country of origin, at a quality bar that makes the obvious landmarks feel shallow by comparison. Getting that to feel human — not like a Wikipedia summary read aloud — took months.

## What We Learned

Scale is a product decision, not just an infrastructure one. When ElevenLabs is the voice delivering every guide in every language to every visitor from 158 nationalities, the cost of audio generation is the cost of the product itself. Every new city, every new language, every personalization pass runs through that pipeline. Optimizing quality, latency, and cost at that layer isn't a nice-to-have — it determines what we can ship and how fast.

informative. It's making something feel alive. That gap — between accurate and memorable — is where ElevenLabs voice quality becomes the difference.

## Traction

Rome only. Zero marketing spend.

  • ~50,000 users acquired organically — starting from 5 posters in 5 churches
  • 50-minute average session (the longest in the travel app category we've seen)
  • 158 nationalities represented in the user base
  • Vatican partnership: Official audio partner for the Giubileo dei Giovani — the Vatican spontaneously directed pilgrims to Cicero
  • Ministry of Culture (MIC): Active partnership to digitize 200+ cultural routes and itineraries across Italy
  • €600K pipeline with Italian Ministry of Culture and Vatican Museums patronage

The ElevenLabs credits would directly fund the audio infrastructure behind every one of these interactions — and behind every new city we unlock next.

Built With

Share this project:

Updates