Inspectial

Culvert and drainage inspection · Pilot open

Culvert inspection software that writes the condition record for you.

Photograph the inlet, barrel and outlet. Inspectial's AI codes debris, sediment, cracking, corrosion, joint defects and scour into a structured, evidence-linked culvert record. Your inspector approves every finding, and the data lands in your GIS, not in a PDF.

Capture offline. AI codes each defect. A human approves. Export to ArcGIS, QGIS or your asset system.

See an example record
  • Works offline
  • Every finding linked to its photo
  • A qualified reviewer approves
  • GeoJSON, CSV, PDF, API

Every culvert photo becomes a coded record your team can review.

An inspector in a hi-vis vest and hard hat crouches beside a concrete culvert and photographs its circular barrel opening with a phone. Water runs out over the invert onto a gravel apron.
Culvert C-104244.9212, -93.4687 · culvert pack
  • Headwall · surfaceweatheringAI suggested
  • Barrel · surfacestainingAI suggested
  • Barrel · vegetationmoss at wall baseAI suggested
  • Outlet · flowclearHuman verified
  • Channel · sedimentloose stones (possible)AI suggested
  • Wing wall · surfaceno defects visibleHuman verified
Evidence IMG_4821, IMG_4822 · Confidence 0.91
{
  "asset": "C-1042",
  "location": { "lat": 44.9212, "lon": -93.4687 },
  "headwall": {
    "surface": "weathering",
    "status": "ai_suggested"
  },
  "barrel": {
    "surface": "staining",
    "vegetation": "moss at wall base",
    "status": "ai_suggested"
  },
  "outlet": {
    "flow": "clear",
    "status": "human_verified"
  },
  "channel": {
    "sediment": "loose stones (possible)",
    "status": "ai_suggested"
  },
  "wing_wall": {
    "surface": "no defects visible",
    "status": "human_verified"
  },
  "evidence": ["IMG_4821", "IMG_4822"],
  "confidence": 0.91
}
The numbered boxes on the photo map to the observations in the record, and every value links back to the evidence it came from.

The culvert visit is the quick part. The coding is not.

A single culvert comes back as a pile of loose evidence. Someone in the office then sorts it, matches it to the culvert ID, codes each defect against your rating manual, writes the report and updates the GIS.

Multiply that by hundreds of structures on a route, and the condition data your asset system needs is the last thing to get done. What comes out is usually a PDF or a spreadsheet that someone has to re-key.

  • Photosabout 35 per culvert
  • Video clipsinlet, barrel, outlet
  • Voice and text notesunstructured
  • GPS and the last inspectionin separate places
Then, by handcoding against the manual, report writing, GIS updates

How it works

One workflow from the site visit to your asset system. AI does the coding; people make the call.

  1. Capture offline, culvert by culvert

    The field app lists the culverts on today's route by distance and walks the inspector through inlet, barrel, outlet and roadway. Every photo, clip and voice note is stamped with time, GPS and culvert ID. No signal needed.

  2. AI codes the defects

    Evidence is grouped by component and turned into coded observations: debris, sediment, cracking, corrosion, joint defects and scour, each with a severity, a confidence and the photos it came from.

  3. An inspector approves

    A qualified reviewer checks each finding next to its photos and approves, edits or rejects it. The AI's grade is kept separately from the approved grade, so you can always see what a person changed.

  4. Export to your GIS

    Approved culvert records go out as GeoJSON, CSV, a PDF report or through the API, keyed to your culvert IDs and ready for ArcGIS, QGIS or your asset management system.

What the culvert pack records

Four components, fifteen observation types and a four-step condition scale where the worst defect controls. During the pilot we map it to your agency's rating manual and culvert IDs.

Inlet and outlet

  • Debris blockage
  • Sediment
  • Vegetation
  • Erosion
  • Scour / undermining
  • End structure damage
  • Perched outlet
  • Piping / seepage

Barrel

  • Cracking
  • Spalling / exposed rebar
  • Corrosion
  • Deformation
  • Joint defect
  • Misalignment
  • Sediment
  • Debris blockage

Roadway and embankment

  • Roadway distress
  • Erosion
  • Piping / seepage
  • Vegetation

Condition scale

Good
Functions as designed. Minor, isolated defects at most.
Fair
Moderate defects that don't yet threaten structure or flow. Monitor or schedule maintenance.
Poor
Significant defects affecting structure or hydraulics: section loss, open joints, active scour. Repair needed.
Critical
Failure or imminent failure: collapse, perforation with soil loss, full blockage. Immediate action.

Reference standards

  • AASHTO Culvert and Storm Drain System Inspection Guide
  • FHWA Culvert Inspection Manual
  • State DOT culvert rating manuals
  • Your own condition codes

Capture in the field. Review from any desk.

Field app preview: a map of today's route with markers by inspection status, working offline, with C-1046 selected and ready to inspect. Web app preview: the Route 12 drainage project on a map, with geotagged field photos pinned at each site and a sidebar listing each site's review status.

Field app (iOS and Android) and review web app. In development.

AI creates the inspection record, not just the report

Keep your GIS and asset system. Inspectial automates the inspection-coding layer in front of it, under rules you can trust.

Agent-ready data, not documents
The structured dataset is the source of truth, so your GIS, scripts and AI agents read it directly. Reports are generated from approved observations, never written straight from the media.
Evidence first
Every observation references the photos, clips or notes behind it. A finding never exists only as an AI sentence.
A human approves
AI proposes; a qualified person decides. Inspectial is not an autonomous inspector, and only approved inspections become official records. Every value is tagged with who stands behind it.
Location is first-class
Evidence, observations and assets carry coordinates, so the output drops onto a map without cleanup.
Offline-first
The full capture flow works with no signal. Data syncs when the connection returns.
Configurable, not custom-built
Your checklist, condition codes and report format live in a template we set up with you. Not a new app.

Who it's for

Drainage and culvert inspection contractorsYou already have crews, cameras and reporting workflows. Inspectial is built to cut the hours spent coding photos and writing reports, so the same team covers more structures.
Counties and municipalities with small GIS teamsGet culvert condition data that lands in your GIS, keyed to your asset IDs, without anyone re-keying it from PDFs.
Road maintenance contractorsInspect long runs of cross-drains along a route, with status and priority per culvert on one map.

Outputs and integrations

Inspectial feeds the tools you already run. It doesn't ask you to move your asset data anywhere, and because the output is structured, evidence-linked data rather than a document, it also works as input for AI agents and automations.

Works with

  • ArcGIS
  • QGIS
  • PostGIS
  • MapLibre

Formats

  • GeoJSON
  • CSV
  • PDF
  • REST API
  • Excelplanned
  • Shapefileplanned
  • GeoPackageplanned

Questions

Something else? Write to pilot@inspectial.com.

Does the AI replace the culvert inspector?

No. AI drafts the coded observations; a qualified inspector or engineer approves, edits or rejects each one. Only approved inspections become official records, and every value shows whether a person verified it.

Can it follow our state's culvert rating system?

Yes. The culvert pack is a template: components, defect types and the condition scale are configuration, not code. During the pilot we map it to your manual and your existing culvert IDs.

What if the photos don't show the defect?

The AI can mark a culvert or component as not assessable instead of guessing, for example when the barrel is submerged or the inlet is overgrown. The reviewer sees that and can ask the field team for more evidence.

Does the field app work without signal?

Yes. The full capture flow works offline, in a ditch or under a road. Photos, clips and notes sync when the connection returns.

How does the data get into ArcGIS?

Approved inspections export as GeoJSON or CSV with coordinates and your culvert IDs, or through the REST API. ArcGIS, QGIS and PostGIS read them directly; there's no proprietary format to convert.

Can we use our past inspections?

Yes, that's how the pilot starts: bring historical photos, video and the reports you wrote from them, and we compare the AI's coding with your inspectors' findings.

Join the culvert inspection pilot

We're onboarding a small number of drainage and culvert inspection teams. Tell us what you inspect, roughly how many structures a year, and which GIS or asset system you report into.

  • Bring real inspections: photos, video and the reports you wrote from them.
  • We set up a culvert template that matches your existing condition codes.
  • You review the AI's structured output against your own findings.
  • Exports arrive as GeoJSON, CSV or PDF, or through the API, in the format you need.