Start where effort repeats.
Study return visits, asset reconciliation, drawing correction and report rebuilding together. Measure the entire accepted deliverable, including the difficult cases.
RESEARCH & PRODUCT DECISIONS
Build the advantage
into the work.
We study the established inspection platforms, capture methods and information standards, then turn the findings into specific product decisions and tests.
OUR WORKING THESIS
Reduce the effort between
evidence and accepted work.
Drawing preparation is a promising starting point. We have not established that it is the largest cost in every inspection program.
Make the record reusable.
Connect capture, identity, inspection locations, revisions and reviewed outputs so the next job starts with an understood history.
This is our product thesis, not a demonstrated competitive advantage or savings claim. The notes below distinguish published facts from our design implications.
PRIMARY SOURCES / REVIEWED 5 OCTOBER 2026 UTC
What we learned.
What we will test.
Vendor documentation shows stated capabilities. Standards and technical guidance inform requirements. Neither is evidence of INSP Works field performance.
8 research notes shown.
01 / Inspection & migration
The market already joins drawings to inspection work.
Cenosco’s May 2026 release notes describe drawing-linked CML selection, equipment uniqueness and improved maintenance-system alignment.
Cenosco 2026 product update
Vendor documentation / 8 May 2026
Design implication
A useful drawing-to-report experience must work through actual asset and inspection records, not stop at a model viewer.
Evidence we still need
Compare location selection, corrections and report handoff on the same bounded work pack.
02 / Inspection & migration
System transition includes a service and data problem.
Antea describes asset data registration, migration utilities, API connections and continuing data-management services alongside its platform.
Antea platform
Vendor documentation
Design implication
Plan data cleanup, historical access, exception review and support as part of adoption. A software screen does not complete migration.
Evidence we still need
Reconcile a real authorized export, then test a parallel run and a documented return path.
03 / Inspection & migration
Inspection scope is wider than recording a picture.
API describes API 570 as covering in-service piping inspection, rating, repair and alteration. Its fifth-edition announcement also describes inspector and piping-engineer roles.
API 570 announcement
Standards publisher / overview, not the complete code
Design implication
Keep capture and record assembly distinct from the qualified assessment and program-specific acceptance.
Evidence we still need
Ask the inspection lead to define required records, roles and the governing program before designing its workflow.
04 / Capture & geometry
Reconstruction needs useful views and a checked result.
COLMAP documents image-based reconstruction from photographs through camera recovery and sparse or dense geometry. Its FAQ discusses weak texture and problems caused by holes and outliers.
COLMAP tutorial
Primary technical documentation
Design implication
Preserve source views and coverage gaps. Compare established engines before developing a new reconstruction engine.
Evidence we still need
Use a representative area, include difficult surfaces and record failed or incomplete regions.
05 / Capture & geometry
A device specification is not the project acceptance test.
NavVis’s accuracy-verification publication describes a survey-based evaluation approach with reference systems, outliers and results.
NavVis accuracy-verification overview
Vendor methodology overview / September 2024
Design implication
Agree independent checks for the intended output and site conditions. Do not turn a device claim into an INSP Works accuracy claim.
Evidence we still need
Record reference measurements, coordinate basis, required locations and the limits accepted by the recipient.
06 / Capture & geometry
A streaming format is one part of interoperability.
OGC’s 3D Tiles standard addresses streaming and rendering large 3D datasets, including point clouds and photogrammetry. It does not establish inspection identity or measurement acceptance.
OGC 3D Tiles
Standards organization / published standard
Design implication
Choose interchange formats and versions deliberately, while preserving the original records and semantic mappings.
Evidence we still need
Round-trip a representative dataset and check units, coordinates, identities, missing regions and editable deliverables.
07 / AI & usability
A fluent answer needs evidence and evaluation.
NIST’s Generative AI Profile addresses risks including fabricated content and information integrity, with proposed measurement and governance actions.
NIST Generative AI Profile
Public technical guidance / July 2024
Design implication
Start with retrieval, comparison and drafting. Evaluate citations, missing information and access boundaries before a live assistant can affect work.
Evidence we still need
Use conflicting, incomplete and out-of-scope records; check whether each answer is supported and appropriately bounded.
08 / AI & usability
The interface has to work where the work happens.
W3C’s WCAG 2.2 guidance covers visible keyboard focus and usable target sizes, among other accessibility requirements.
W3C WCAG 2.2
Web standard
Design implication
Make the main journeys usable by keyboard and on narrow screens. Treat accessibility checks as part of each release.
Evidence we still need
Exercise real controls, downloads, error states, focus and reflow; a checklist alone is not conformance.
THE NEXT EXPERIMENTS
Turn the research
into a measured advantage.
Each workstream needs a deliverable, a comparison and a reason to keep going.
Capture to drawing
Compare methods on the same representative area. Measure coverage, location errors, corrections and time to an accepted drawing.
Evidence to report
Measure source retrieval, draft preparation, review effort and returned corrections using a client-defined output.
Existing data to daily use
Test import reconciliation, historical access and a repeatable handoff before making a system-transition decision.
PROOF THAT CAN BE REPEATED
Keep quality and cost
in the same comparison.
Our proposed study compares equivalent scope and accepted output. Record unsuccessful capture, missing locations, repeat visits and reviewer corrections. Separate first-job setup from later-job assumptions and retained staff capacity from realized cash savings.
The website’s example deliberately allows assisted work to take fewer hours and still cost more. The correction scenario shows how additional labour changes the result. These are arithmetic examples, not commercial forecasts.