Deloitte demos AI tool to speed pedestrian‑safety planning, cites Caltrans use

Deloitte demonstrated an AI-enabled tool its team said can compress months of pedestrian‑safety analysis into hours while surfacing data and provenance for agency review.

Josh Nisbett, Deloitte’s managing partner for state and local government in Arizona, introduced Vikram Manda, lead for Deloitte’s transportation practice, who showed a demo of ‘Infrastructure Insights Pro.’ Manda said the platform ingests agency datasets — historical crash records, project‑management systems and GIS layers — and provides a GenAI question panel beside an interactive map so users can zoom to districts and inspect collision points.

Manda described a built‑in “data trust score” that ranks and documents data provenance so agencies can judge the reliability of AI‑generated recommendations before acting. In the demo environment he used sample data and noted results will vary with a department’s full dataset.

Manda walked the committee through a use case for identifying vulnerable road users and producing a concept report. He said the traditional, manual process of analyzing crashes and drafting a concept report can take roughly six to eight months; the platform can produce a solid draft in minutes or a few hours that safety engineers can refine. He cited an implementation for Caltrans that, he said, yields near‑final reports containing analysis, maps and supporting imagery.

Committee members asked whether the tool has been used in other states and whether it produced cost savings. Manda replied that the safety use case is from Caltrans, that Deloitte is implementing related use cases in multiple states, and that the primary measurable benefit is reduced staff effort — which he equated with cost avoidance — rather than a single dollar savings figure presented to the committee.

The presentation included dataset excerpts (Manda noted a demo figure that in the sample showed roughly three‑quarters of fatalities in that dataset were bicyclists) and an explanation that results depend on data completeness. Manda emphasized that the tool produces drafts and insights to aid human safety engineers, not replace their judgment.

The chair thanked Deloitte for the briefing and noted the presenters would appear before a related AI committee the following day.