Executive Summary
California’s state highway network generates massive volumes of traffic, crash, and safety data.
For years, that data lived in multiple separate systems. Engineers spent hours assembling traffic
and safety pictures before analysis could begin, and safety investigations started only after
crashes occurred.
Under Governor Newsom’s Executive Order N-12-23 on Generative AI, the California Department of
Transportation brought two production generative AI platforms into service. Traffic Mobility
Insights (TMI), launched August 2025, unifies seven statewide traffic systems and nearly 1,500
vetted policy and operational documents into a single geospatial platform where engineers query 10
terabytes of roadway data and 33 billion mobility records in plain English. Road Safety Insights
(RSI), launched November 2025, integrating more than 30 datasets to identify high-risk locations
for pedestrians and bicyclists. One in four fatal or serious-injury crashes in California involves
someone walking or biking.
TMI reduced statewide corridor analysis from hours to minutes, with 95 percent accuracy for
station-level delay detection. In RSI’s first two weeks, safety engineers across all 12 Caltrans
districts generated 560 draft safety investigation digests. Each previously required 20 to 30 hours
of manual compilation; the two-week total represents an estimated 11,200 to 16,800 hours of manual
work avoided. Outputs cite published standards.
Both platforms serve headquarters and all 12 districts. Each completed risk and privacy review
before deployment, and staff review every output before it informs a decision. Caltrans treated
these as production systems, not experiments.
