Skip to content
The product

Capture to controlled issue, without leaving the platform

6 connected capabilities that take a raw drone dataset all the way to a defensible, delivered report. Here's how each part works — and how they fit together.

The lifecycle

Five gated stages, one state machine

The workflow is enforced, not suggested. An inspection can't skip a gate — it moves forward only when the work for the current stage is genuinely complete.

01
INS-01

Capture & ingest

Upload thousands of drone images in one resumable session. EXIF, GPS and capture time are extracted automatically, and web + thumbnail derivatives are generated so the whole set is browsable in minutes — not after an overnight copy.

02
INS-02

Capture QA

Every image is checked for coverage and quality before analysis begins. Gaps and unusable frames surface as named issues, so a report is never built on a set that quietly missed an elevation.

03
INS-03

AI-assisted analysis

Run AI defect detection across the set on demand. Suggested defects land in their own review queue as annotated regions — corrosion, cracking, lifted laps, blocked outlets — for an inspector to accept or reject. Nothing reaches a client unreviewed.

04
INS-04

Technical review

A qualified reviewer confirms findings, sets severity and recommendations, and clears completeness gates. Author and approver are distinct, and the report can't advance while required scope, method or limitations are missing.

05
INS-05

Controlled issue

Approve to issue and the report is frozen: numbered, checksummed and immutable. Re-issues are tracked as new controlled versions, so there's always a defensible record of exactly what was delivered, and when.

ai defect detection

AI-assisted defect detection

Point the AI at an inspection and it works through the imagery, flagging visible defects — corrosion and cut-edge corrosion, cracking, holed or lifted sheets, open laps, loose fixings, moss, debris and ponding — as normalised regions with a class, confidence and description.

Every suggestion is advisory. It sits in a review queue keyed to the exact image and only becomes part of the record when an inspector accepts it, at which point it converts to an ordinary numbered defect. A high-detail tiled pass finds finer detail on busy roofs, and duplicate detections across overlapping photos are merged so the same crack isn't counted twice.

Try it free
  • Replaceable providers — cloud vision today, self-hosted models when you want them
  • Suggestions are scoped, confidence-ranked and inspector-verified before issue
  • Standard and high-detail tiled passes; automatic duplicate merging
  • Real per-run cost and credit tracking, with a per-job spend ceiling
annotation

On-image annotation

Mark up the evidence where it lives — on the image. Boxes and polygons are drawn straight onto the photograph and bound to a defect record carrying a stable reference, type, severity and recommended action.

Accepted AI suggestions reuse the exact same annotation path, so an AI-found defect and a hand-drawn one are indistinguishable in the final register — one consistent, reviewable defect model throughout.

Try it free
  • Pan-and-zoom viewer built for large, high-resolution drone frames
  • Stable D-NNN defect references with severity and recommendations
  • AI and human findings share one defect model
  • Every change audited
controlled reports

Controlled condition reports

The report is a controlled document, not a loose export. On approval it's frozen and issued with a number and a content checksum, so a delivered report can always be verified as genuine and unaltered.

Completeness gates hold the report back until required scope, method and limitations are present. Where AI assisted the analysis, the methodology carries a clear, inspector-verified advisory note. Re-issues become new controlled versions rather than silent edits.

Try it free

Roof condition report

INSP-2026-014 · Rev 1

Issued · frozen
D-001 Critical

Cut-edge corrosion, north slope

Recommend localised treatment within 3 months.

D-002 Major

Ponding at central outlet

Clear outlet; monitor after next rainfall.

Issued 14 Feb 2026 SHA-256 · 4f7a…e19c

Representative preview — layout only. Wire a real issued PDF here when available.

resumable upload

Resumable large-dataset upload

Drone jobs mean thousands of large files, often from a site with patchy connectivity. Uploads are chunked and resumable: a dropped connection doesn't restart the transfer — it continues from the last confirmed chunk.

Chunks are assembled server-side by streaming, identical files are de-duplicated within an inspection, and abandoned sessions are swept up automatically. Big sets go up reliably, without babysitting.

Try it free
  • Chunked transfer with automatic retry and resume
  • Content de-duplication within an inspection
  • Streamed server-side assembly — never whole-file-in-memory
  • Abandoned sessions reaped automatically
multi tenant

Multi-tenant isolation

The platform is multi-tenant to the core. Every record is stamped with its tenant on write and filtered by tenant on read, enforced centrally so a query can't accidentally reach across boundaries. Cross-tenant writes are rejected outright and a request with no tenant context fails closed.

Storage paths are tenant-scoped, share-link access resolves strictly to the owning tenant, and Enterprise can take a dedicated database. Isolation is covered by a continuously-run test matrix, not left to convention.

Try it free
  • Tenant stamped on write, filtered on read — enforced centrally
  • Cross-tenant access rejected; no-context requests fail closed
  • Tenant-scoped storage and share-link resolution
  • Dedicated-database option for Enterprise

See it on your own imagery

The fastest way to judge an inspection platform is to run a job through it. Start free and put a real dataset in front of the AI, the review queue and the report.

14-day free trial on every plan. No card required. Professional-tier features, 100 AI credits, 50 GB and watermarked reports during trial.