Status
Bus Safety has been in production since November 2025, used by around 1,000 drivers at a Japanese bus operator. The next step is exploring rollout to additional bus operators.
Daily safety scores for bus drivers
Summary
At the bus operator where it has been deployed, accident frequency improved year-on-year from 19% to 17%, and loss ratios improved by around 10%.
The problem
As bus drivers age and fewer new drivers join the industry, an AI scoring system is helping make safety coaching more effective by surfacing risky driving behaviour from daily driving data. It analyses 69 driving-behaviour features captured from dashcams, using our proprietary AI scoring model to produce a risk score for each driver, along with a clear explanation of what drove it.
Solution
We built a scoring algorithm designed specifically for route buses. It extracts 69 driving-behaviour indicators from dashcam data and feeds them into our proprietary AI scoring model to produce daily and monthly risk scores.
Bus safety coaching has relied heavily on individual instructors' experience and instinct, and an ageing workforce means fewer experienced instructors are available. At the same time, bus insurance loss ratios remain stubbornly high, making accident prevention and safe operations a pressing priority. Existing telematics scores weren't built for large buses, and couldn't assess behaviours specific to route buses, such as door operation or indicator use. Without a way to standardise coaching, it's hard to raise safety levels across a workforce with limited instructor capacity.
Thanks to its interpretable ("glass-box") design, we can show each driver exactly which behaviours affected their score, making feedback specific and easy to act on. Scores are available as soon as a shift ends, so drivers can reflect while the details are still fresh.