Read every trip
An in-cab AI unit streams GPS, hard braking, sharp turns and time of operation from every cab, in real time.
所有案例研究Electric-vehicle fleets · JOIE · Tai Wo Management
Commercial motor, priced on how each machine actually drives.
JOIE is putting 1,000 electric and hybrid cabs on Hong Kong's roads, run by Tai Wo Management. Each one carries an AI unit that streams its driving to YAS. YAS scores how every driver actually drives, and a licensed insurer prices each driver on that score — Hong Kong's first taxi insurance based on behaviour, not averages.
1.28M
trips scored
17M km
commercial driving scored
364
taxis actively scored
1st
behaviour-based taxi fleet in HK
合作伙伴
Underwritten by Zurich · scored by YAS (ARIA engine)
缺口
Commercial motor was priced for a world that's already gone.
Taxi insurance is still set once a year, on variables that have nothing to do with how a fleet actually drives — and that single cadence costs operators on both sides of the ledger.
Before YAS
Priced once a year
Renewal premiums locked to vehicle age and demographics. Nothing a driver does moves the price until the next renewal.
Safe driving unrewarded
Everyone pays the fleet average, so careful drivers subsidise reckless ones.
Real cost stays invisible
Claims driven by behaviour the old policy never measured.
With YAS
Live score per driver
Every trip updates a behaviour-based risk score, visible to operators and insurers.
Insurer-supported pricing
Licensed carriers use the score to support pricing; authority stays with the insurer.
Coach before claims
Risky behaviour flagged early, so fleets improve before losses land.
方案
An in-cab AI reads every trip; AURA scores every driver.
Every cab streams its driving to YAS. The ARIA engine scores each driver, and a licensed insurer sets the price from that score. The platform now shown on our platform pages was built for JOIE first — and it has since scored 1.28 million trips on about 8.9 TB of driving data.
Score with ARIA
A new driver starts on the fleet average; after about 30 scored trips their own record takes over and the price becomes personal.
Price with GLM
Poisson frequency × Gamma severity, validated against confirmed loss — the insurer prices off the score.
Reward & steer
A live dashboard, driver rankings and safe-driving rewards turn the policy into a programme the operator can steer.
“Our mission isn't just to swap out petrol vehicles — it's to lead the industry toward smart urban mobility.”
Traction & risk trend
Ten times the distance in four quarters — measured the whole way.
JOIE scaled fast: trips up 8.6×, distance up 10×, and the active fleet up 3× inside a year. As that many new drivers joined, the fleet-wide average eased back a few points — the arithmetic of a tripling fleet, not tenured drivers getting worse; new joiners enter below the line they soon climb past. What didn't move was the measurement: the same framework showed exactly which factors shifted and by how much, with speeding the consistent one to coach.
Q3 2025
- 行程
- 52,386
- 距离
- 0.6M
- 车队分数
- 77.0
Q4 2025
- 行程
- 285,860
- 距离
- 3.6M
- 车队分数
- 75.9
Q1 2026
- 行程
- 486,422
- 距离
- 6.7M
- 车队分数
- 74.9
Q2 2026
- 行程
- 452,661
- 距离
- 6.1M
- 车队分数
- 73.5
Fleet score is the average across all active drivers; the slide reflects a 3× larger, younger driver pool, not tenured drivers regressing.
Impact on driving
How we got the fleet driving safer.
The same scores that price the policy power a live driver leaderboard — every driver sees exactly where they rank. Making safe driving visible and rewarding it gives each one a reason to climb, with a clear target to coach: speeding, the factor that moves the score most. The careful driver moves up; new joiners start below them and work up the same table.
What makes the score worth chasing is that it's their own. A new driver starts on the fleet average; credibility weighting then shifts the price onto their personal record by around 30 trips. Drive better and your number moves, fast.
62% of the fleet
Carries 0.45× base claim frequency — the safest drivers, priced accordingly.
18% of the fleet
Preferred. Consistent behaviour earns a modest discount.
12% of the fleet
Watch band — elevated frequency; coaching flagged.
8% of the fleet
Carries 2.40× base frequency — priced for the risk and reviewed closely.
成果
Proven at fleet scale. Ready for what moves next.
JOIE isn't a one-year loss-ratio story. It's a method proven at scale: 1.28 million trips, about 8.9 TB of driving data, and the same model already reading fleets beyond JOIE. The engine that breaks a taxi fleet's risk down factor by factor is the one we now point at autonomous vehicles, robots and drones.
~8.9 TB
calibrated telematics, and growing
10×
driving distance in a year
Beyond JOIE
the same model reads external fleets