Auditable evidence for every demolition, ready without rebuilding it

The problem

The evidence is rebuilt by hand, and by then the building is gone.

Quality and safety audits, asbestos inspections and document handovers to developers demand evidence that today gets rebuilt by hand. The more visible the project — hospitals, stadiums, public infrastructure — the higher the reputational risk of a non-conformity from documentary gaps.

• An inspector, the site supervisor or the developer asks for the evidence of an asbestos removal or of a specific demolition phase. Nobody refuses — somebody simply spends days assembling it.
• It gets assembled from loose photos scattered across several phones, paper forms in a folder in somebody's van, and whatever the foreman who happened to be there that week still remembers.
• The structure is already down, so anything not captured at the time cannot be captured now, at any price. That is what makes a demolition audit different from an audit in a factory that is still standing.
• And the more visible the project — a hospital, a stadium, a piece of public infrastructure — the more a non-conformity from a documentary gap costs in reputation, well beyond the file it appears in.

How it fits the IRIS system

Edge captures in the moment; Agents organizes it all by the applicable standard.

Edge automatically captures a pre/post photo of each critical point and the structured log of decisions and sign-offs. Agents organizes all that evidence by the applicable standard, ready to download from the web with no manual post-hoc work.

The direction is what matters. A document manager stores whatever somebody decides to upload afterwards, and so it inherits exactly the same gaps as the manual assembly. Here the dossier closes the instant the phase happens, on the site, and what Agents does later is reindex that same evidence for whichever standard or whichever party happens to be asking.

See the full IRIS architecture →

Before and after

Today's rebuilt evidence versus the dossier that builds itself

Aspect	Today	With iLEAN Agents
Preparing a project handover	Days of rebuilding by hand	Hours of review, at most
Pre and post photos	Scattered across several phones	Attached to the phase they belong to
Decisions taken on site	In somebody's memory, if anywhere	A structured, timestamped log
Sign-offs	Signed forms in a folder	Timestamped and indexed
A missing piece of evidence	Found by the auditor	Flagged when it should exist
For a different standard	Rebuilt from scratch	Same evidence, another index

Before: days rebuilding evidence with non-conformity risk. After: automatic dossier, zero manual post-hoc work.

Impact estimate

Impact estimate — to be validated with your numbers.

The block below is an estimate to be validated against your plant's actual data. We put it forward so the committee has an order of magnitude; we refine it during the assessment.

• Estimated payback 6-12 months, depending on how often you are audited and how many project handovers you close in a year.
• Days of rebuilding evidence from phones and folders replaced by zero manual post-hoc work.
• Protection of the authorizations and certifications your contracts depend on, the asbestos-work authorization included.
• And on hospitals, stadiums or public infrastructure, the reputational risk of a non-conformity from a documentary gap drops to what was actually captured on site — which is everything.

Estimated payback of 6-12 months depending on audit/handover frequency, protecting authorizations and certifications. Estimate to validate.

And the fair question from the production manager

"Isn't this just another document manager?" — the difference is when the dossier closes, and on a demolition that is the only thing that matters. A document manager stores what somebody uploads after the fact, so it inherits every gap the site left behind; here the phase generates its own evidence while the structure is still standing. Organizing that evidence against a standard is an anchored task — a closed set of requirements against a closed set of records — where the best models drop below 1.5% error [1], and every sign-off inside the dossier was signed by a person, not generated.

^[1] OpenAI paper "Why Language Models Hallucinate", 2025 — on the reliability of AI in anchored tasks.