Tokyo HQ · April 2026 · confidential
Thirty-five million travellers visit Japan a year. The world's largest booking platforms cannot see most of the inventory they actually want, cannot speak the language the hosts actually speak, and cannot pick up the phone when the trip breaks.
Net Promoter Score for international travel booking in Japan: –18. The 89-point trust gap, applied to one country.
The entire accommodation + trip layer. The pond we swim in.
35.3M inbound travellers · avg spend €1,360 per trip · 2024 baseline, JTB figures.
4.5M trips/yr with ≥4 travellers or ≥7 nights · avg €1,350 · €210 fee per completed trip.
In 2025 an agent stack could finally translate, negotiate, re-book, and route — in production, at SLA, below human cost. The twelve-minute replace was fiction two years ago.
Google's SGE and ChatGPT browsing collapsed the SEO-bought top of funnel that aggregators depended on. Travellers ask an agent, not a URL bar. The incumbents are out of position.
Yen weakness drove 35M visitors in 2024. Simultaneously, prefectural minpaku enforcement is cancelling tens of thousands of listings a quarter. Trust premium has never been higher.
Six agents. One shared inbox. One phone number that picks up in the prefecture you're standing in.
We do not sell a feature list. We sell the morning in Gion where nothing goes wrong — and, when it does, the twelve minutes during which the family never knows.
walked within last 180 days
minpaku + overbooking + other
from call to a human in your prefecture
Not feature-level. Structural. Each of the following would require Booking.com to rewrite something they are not organisationally able to rewrite.
Independent ryokans, machiya, and family-run inns are sourced prefecture by prefecture, on foot, in Japanese. Booking.com's entire supply engine is globalised, English-language, and contract-based. Replicating Tabi's inventory requires them to re-build their BD motion as a local one — which they have no incentive to do because it cannibalises their existing chain-hotel commission revenue.
Twenty-four humans, six pods, each answering calls in the prefecture the caller is standing in. Booking.com's call-centre model is globalised by design — they cannot build same-prefecture pickup without scrapping a twenty-year operational thesis. Every minute of <60s regional pickup is a compounding trust gap.
Tabi is paid €210 when a trip is delivered; €0 when it fails. Booking.com earns 15–25% commission on the property. To match our pricing, they would need to invert the revenue model that funds their 27,000 employees — a structural move that equity markets would punish before we felt it.
Built Mercari's cross-border unit from zero to 4M users. Grew up between Osaka and Vancouver — knows both edges of the trust gap personally.
Shipped Stripe's Japan payments stack. Ran the agent infrastructure team that handled Stripe's first production LLM workloads in APAC.
Product lead on Anthropic's agent platform. Writes the eval suite before the agent ships. Author of the "taste as a product metric" internal memo.
Ran airport recovery operations for ANA for eleven years. Knows how to get a family of four into a hotel at 10am when everything has just broken.
Led by SoftBank Vision Fund. 18 months to national coverage and the first adjacent market.
All six agents in production. Replace-in-place hits 12-min SLA at 95%+. Language Bridge live in Korean or Thai — not only EN ↔ JP.
€10M ARR. 80,000 completed trips. 65% 6-month family repeat. 2,000 verified properties.
Six JP pods fully staffed · Tokyo HQ at 50 · 98% Verification coverage · first adjacent pod (Seoul or Bangkok) live with 5 humans + 200 properties.
WTTC 2025 figures — the pond, not the pool.
Distance between NPS for legacy online-travel platforms and the best-in-class traveller experience. Tabi's current Japan NPS of +74 already closes 92 points of that gap.
Flowed into AI-native startups. Travel is the next vertical where the incumbent is large and structurally slow.
Approx. 27,000 employees. Revenue-per-employee ~$860k. We run at €1.9M per employee today — at 3% of the headcount, by design.
JNTO figures. Projected to exceed 40M in 2026 under the government's visitor target.