June 19, 2026, (Inside AI) — The National Highways Authority of India has deployed an in-house artificial intelligence system to flag faults in Detailed Project Reports and track road defects, officials said.
The system, built on NHAI’s vast repository of circulars, acts, Indian Roads Congress codes, and monthly reports, serves as an assistance tool to reduce the burden of sifting through hundreds of pages of guidelines.
NHAI Chairman Santosh Kumar Yadav said the tool emerged from a workflow study that found project directors and regional officers were bogged down by tasks easily handled by AI. Access is currently limited to NHAI officials, but Yadav confirmed it will extend to consultants and contractors after security vetting.
Photo Uploads Reveal Road Defects Instantly
Field engineers can now photograph a defect and receive immediate analysis. The system scans guidelines, codes, and specifications to deliver results in moments.
Read: Hong Kong Hospital Deploys Privacy-First Patient AI on AWS
“This system was needed because when we carried out the workflow study of Project Directors and regional officers of NHAI last year, we found that they have to be unnecessarily involved in a number of work, which can be easily done through AI tools. With this system in place, any official can check the defect of a highway by just uploading a photo. The system will quickly run through hundreds of guidelines, codes, notification, specifications and give the result,” said Yadav.
Tackling the DPR Problem at Its Root
Faulty DPRs have long plagued highway projects. Union Minister Nitin Gadkari has repeatedly blamed poor civil engineering and substandard reports for rising accidents and fatalities.
Yadav noted that a Rs 2,000 crore project often sees a Rs 200–300 crore variation. Now, officials can upload a DPR and the AI will find errors against IRC codes and policy circulars, allowing fixes before construction begins.
A Technical Schedule Analyzer tool reviews DPR’s Schedule B and C documents—the core technical annexures defining physical scope—and flags findings by severity and completeness.
Margsarthi Chat Agent Draws 50,000 Queries
An in-house chat-based AI agent, Margsarthi, went live on April 18. Connected to NHAI’s datalake, it has processed over 50,000 queries from roughly 1,100 users, with over 40% of questions about circulars and documents.
“When we started working on this, we found that across the organisation, the officials used to keep personal folders of circulars, Acts and similar reference material on their own computers. Even for experienced officers, it was difficult to keep track of the sheer volume of rules and regulations in force. With this system, they can get it in just one query,” said a senior project engineer.
Kick-Off and MPR Tools Surface Bottlenecks
Additional tools, Kick-Off and MPR Insights, track issues from kick-off meetings and monthly progress reports. They flag critical bottlenecks, surface overdue items, and provide recommended actions with direct citations to source pages.
Read: India Targets 3-Year Justice Cycle with AI Predictive Policing
The deployment marks a shift toward AI-driven quality control in India’s highway sector, though questions remain about data security and algorithmic bias as external access expands. NHAI has not disclosed the model’s architecture or training data specifics.
