AI below grade in 2026: utility mapping, sewer inspection coding and tunnelling records

Before you dig: subsurface utility mapping

Exodigo says it maps utilities, foundations and geotechnical conditions without breaking ground. It describes four stages: AI research across utility records, as-builts, geotechnical reports, satellite imagery and permits (ExoInfo); non-intrusive multi-sensor scanning with AI analysis; design deliverables and digital twins; and utility relocation services. Engineers and geophysicists are part of the delivery. It cites $75B of de-risked investment and 86.1M sq ft scanned (vendor claims).

This is our note: the pages we read do not say how results map to ASCE 38 utility quality levels. Owners and designers who specify those levels should ask.

Existing networks: AI coding of sewer CCTV

SewerAI's AutoCode assigns NASSCO observations to CCTV footage and claims 99%+ accuracy at 6x manual speed; Risk & Rehab turns inspections into rehab plans, and Sewer3D models manholes for MACP surveys. ITpipes' AiDetect combines AI with expert human oversight and claims at least 50% less coding and QA time with 97% defect detection accuracy. Both say outputs are NASSCO-compliant.

During tunnelling and underground work

GroundedAI targets mining, tunnelling and engineering consultants: Lithos captures 3D data, photos and notes offline in the field, Strata turns them into shared views and repeatable reports, and data exports to Leapfrog, Deswik and AutoCAD.

Pricing

None of the four lists prices; all route to demos, sales or contact.

What to ask before a pilot

These are our suggestions, not findings. For utility mapping, compare results against potholes or test pits at known locations. For CCTV coding, have a NASSCO-certified reviewer code the same footage and count disagreements on structural defects. For underground records, ask how ground-condition data ties back to design assumptions and change claims.