AI and machine-learning-powered travel intelligence

The AI layer that learns what people actually want to do.

Done It Yet is building a proprietary AI- and machine-learning-powered intelligence layer for activity-led travel. The live system is the foundation: it collects signals, structures fragmented operator supply, and stores the graph state future models will learn from.

The heavy Google Cloud phase comes next. Credits unlock broader signal coverage across social, web, and market data; denser inference and refresh cycles; and proprietary ranking, prioritization, and matching model training on top of the current substrate.

Join the waitlist See the live foundation

Live now

Data-and-decision substrate

What is live today is the foundation, not the finished product: a Google-native system that builds the market universe, scores fragmented supply, stores learning state, and decides what should be refreshed, enriched, or pursued next.

36 live handlers 31 scheduled loops 2,318 route edges 1,248 location edges

Ingests and structures operators, routes, locations, and creator surfaces
Persists learning state across rights, supply, and demand signals
Prioritizes what deserves enrichment, outreach, and coverage next
Runs continuously on Google Cloud and Firebase without daily manual curation

Next unlock

Model training, AI matching, and trip building

1,752 Trips mapped multi-day itineraries already inside the engine

1,035 Operators enriched ranked, enriched, and structured for selection

245 Permission-positive creator records already captured in Firebase

2,318 Route edges graph relationships already structuring travel supply

Why AI wins here

This category is built for an intelligence layer.

The experience is the real demand unit

A traveler often starts with what they want to do, not where they want to sleep: a 5-day Sahara dune-buggy trip, a 4-day Lombok to Komodo boat route, or a 3-day horseback safari in South Africa. That makes activity-led trips structurally better suited to matching, ranking, and learning systems than inventory-led travel flows.

Premium spend follows memory, not inventory

A growing premium segment will happily choose comfort when it matters, but the real splurge is the experience itself. Affluent couples, friend groups, and experience-led travelers optimize around the story they want to live, not the hotel category they want to compare.

Signals sit upstream of booking

Travel intent appears in saves, shares, group chats, creator proof, and planning behavior before it reaches a booking flow. That gives the engine richer upstream signals than destination search alone.

The industry still under-models this

Much of travel still organizes around destinations, accommodation inventory, and manual heuristics. Many multi-day operators remain fragmented and under-distributed. This is exactly where learning systems should compound faster than incumbent workflows.

How the company is built

The foundation now. The product layer next.

Layer 1

Travel demand intelligence substrate

Operator intelligence Route graph Creator graph Rights memory Demand sensing Market prioritization

Layer 2

Activity-first matching and booking layer

AI matching Trip builder Bookable routes OTA surface

The live system is the substrate, not the end-state product. It captures the signals, graph state, and operational memory that future matching and booking flows depend on.

Done It Yet is building the same type of structural advantage strong vertical platforms build first: proprietary understanding of demand, supply, and proof before monetization catches up.

Right now the focus is Layer 1. That is where more data, more state, and more model training create defensible advantage. It is also where Google Cloud credits matter most.

Layer 1 captures and structures the signals future models learn from
Layer 2 applies that intelligence to matching, trip building, and booking
Credits accelerate the move from live substrate to model-driven product

36 live handlers 31 scheduled loops 1,035 operators enriched 245 creator approvals captured

Google Cloud at the core

Google Cloud is where the learning system gets expensive.

Foundation live today

Google-native data and decision infrastructure

Today the system already uses Google Cloud to collect and structure the data, orchestrate always-on refresh cycles, persist graph state, and run ranking and classification workflows. That is the foundation. The next phase is heavier: more sources, more refreshes, more inference, and proprietary model training on top of the same stack, using signals that sit upstream of traditional travel demand systems.

Collects and stores graph, rights, and operator state
Runs always-on refresh and prioritization workflows
Supports ranking, classification, and enrichment loops
Creates the substrate for future model training and trip matching

Firebase Hosting

Public-facing deployment layer for the product surface.

Firestore Native

System state, supply graph, permissions, and learning data.

Cloud Functions Gen2

Serverless intelligence and orchestration workflows already live.

Cloud Scheduler

Always-on refresh jobs driving the intelligence engine forward.

Cloud Storage

Asset and content operations supporting the rights-cleared content layer.

Model training + inference

Ranking, prioritization, and matching workloads that expand as signal coverage and learning density grow.

Foundation to scale

The architecture is live. The expensive phase comes next.

What exists today is the substrate. The next phase is where Google Cloud usage expands materially: more signals, more state, more inference, more training data, and model-driven product logic on top of the same Google-native stack, especially as the models learn from signal types the travel industry still underuses.

Live foundation

What already runs today

1,752 tours mapped and 1,035 operators enriched inside the intelligence base
939 seeded candidates with 2,318 route edges and 1,248 location edges structuring the graph
245 permission-positive creator records already inside the rights layer
36 serverless handlers and 31 scheduled loops already orchestrating the engine

Graph state Rights memory Always-on orchestration

Next build phase

Broader signals and denser learning

Expand from social and web signals into richer operator, tourism, and market data across more routes, entities, and planning surfaces. That increases training data, refresh density, and inference volume.

Signal expansion More refresh cycles Training data growth

Model-driven product layer

Matching, trip building, and pricing intelligence

Train proprietary ranking, prioritization, and matching models on top of the substrate, then apply them to trip building and bookable activity-led travel flows, with pricing intelligence as a longer-term unlock.

Matching models Trip builder Pricing intelligence

How the system compounds

Every loop makes the system harder to catch.

Signals

The engine captures what travelers save, share, search, and pursue.

Smarter focus

Better signals improve which operators, routes, and creators deserve attention next.

Better proof

Better prioritization produces stronger creator proof, trust, and trip legibility.

Better matching

Better proof and structure improve trip building, recommendation quality, and conversion.

The compounding loop Activity-first intelligence

More learning. Better decisions. Better trips.

This is why the system should compound faster than incumbent travel workflows: more signals sharpen focus, better focus improves proof, better proof improves matching, and better matching improves both conversion and the next wave of data collection.

More signals sharpen the next ranking and prioritization pass
Better focus improves creator proof, trust, and trip clarity
Better proof improves matching quality and booking intent
Better matching improves operator economics and future data capture

Founder

Built by someone who understands both markets and systems.

Done It Yet is being built by H. Spierenburg, a solo founder with 3 years of strategy consulting experience at an elite European firm across travel, hospitality, and software markets, backed by BSc and MSc training in economics and econometrics.

The conviction came from the same place demand already starts today: reels, blogs, and group chats. Trying to coordinate holidays with friends and separately book a multi-day boat route around Komodo made the gap impossible to ignore: endless inspiration, no definitive layer, and far too much friction when it was finally time to choose.

The founder perspective is pan-European. The company is designed to combine market insight, product discipline, and AI execution into one thing: activity-first travel at scale.

Market lens Travel, hospitality, and vertical software experience

Pan-European lens Commercial perspective across European travel, hospitality, and software markets

Build plan Referrals first, deeper matching and booking control later

Join the waitlist

Be first to access activity-first trip discovery built around AI.

Leave your email and, optionally, the kind of trip you are trying to build.