Knowledge · Technology
Documents as evidence,
not files in a folder.
The quiet idea behind trustworthy construction intelligence is that a document is evidence for a fact, not just a file to store. This reference explains the shift, why provenance makes every value checkable in seconds, and why treating the fact as the unit of understanding is what stops information being sealed inside PDFs.
01 / Overview
The shift from files to evidence
Most construction software treats a document as a file, its job is to store the PDF and help you find it again. The evidence model treats the same document as support for the facts it contains. An invoice is not a file, it is evidence of a supplier, an amount, a set of line items. A drawing is evidence of dimensions and rooms. A contract is evidence of obligations and dates. The document still exists, but it has been promoted from the thing you store to the thing that proves what you know, and the facts it supports become what the business actually works with.
This sounds abstract until you see what it changes. When facts are the unit and documents are the evidence, nothing is hidden inside a PDF, every value the system shows can be traced to the page that produced it, and the same fact described differently across several documents can be recognised as one thing backed by all of them. That is the foundation the rest of the intelligence stack is built on, the document intelligence reference explains how documents are read into these facts, and the construction intelligence reference explains what the connected facts let a business understand.
02 / Key mechanics
Three shifts the evidence model makes
Each one moves the software from managing documents to understanding what they say, while keeping every fact anchored to its proof.
From storing files to supporting facts
Traditional software stores a document so you can find the PDF later. The evidence model treats the same document as support for the facts inside it, the supplier, the amount, the dimension, so the fact is what the business works with and the document is what backs it.
From folders to provenance
Instead of a value appearing in a system with no explanation, every extracted fact points back to the exact document and page it came from. Checking a number becomes a glance at its source rather than a re-read of the whole document.
From filing to understanding
A plan, a specification, an invoice, a contract and a council approval stop being separate files and become evidence about the same job. The job is understood from the facts they collectively support, not from whichever document someone happened to open.
The middle shift, provenance, is the one a builder feels daily. It is what turns an extracted number from a claim into a checkable fact, and it is what makes any AI suggestion safe to act on, because a value you can trace is a value you can verify in seconds. This is why provenance is not a feature bolted on at the end but the precondition for trusting anything the software reads, a point the AI guardrails reference makes as the evidence-on-every-value rule. The durable record all this evidence builds up over a job's life is covered in construction knowledge.
03 / Best practice
Why the evidence model wins on the bad day
An operator observation makes the value concrete. The evidence model does not prove itself on the day everything goes right, it proves itself on the day something is wrong. A supplier queries an amount, a client disputes a variation, a certifier wants the sheet the frame was built from. In a folder-based system that day is an hour of hunting through an inbox for the document that settles it. In an evidence-based system the fact and its document are already linked, so following the number to its source is a click. The model earns its keep precisely when a builder is under pressure to show why a figure is what it is.
The practical implication for choosing software is to look past the storage and ask about the link. Any product can hold your PDFs. The question is whether the values it shows you carry their evidence, whether you can follow any number straight to the page behind it, and whether the same fact from different documents is reconciled rather than duplicated. Those are the properties of an evidence model, and they are what separate software that understands your paperwork from software that merely files it. How a builder should evaluate that, vendor-neutrally, sits in the document intelligence reference.
04 / FAQ
Common questions.
It is a shift in what the software considers primary. Traditional construction software stores documents, so its job is done once the PDF is filed and findable. The evidence model treats each document as support for the facts it contains, the supplier and amount on an invoice, the dimensions on a drawing, the obligations in a contract, and makes those facts the thing the business works with. The document does not disappear, it becomes the proof behind each fact. The practical consequence is that information stops being sealed inside PDFs and becomes data the rest of the system can reason over, while every piece of that data still points back to where it came from.
Because it decides how expensive checking is, and checking is where the time goes. When an extracted value carries the document and page it came from, a person confirms it in seconds by glancing at the source. When it does not, the reviewer has to re-derive the number or trust it blindly, and neither is cheap. Provenance is also what makes AI output safe to use at all, a suggestion you can trace is a suggestion you can verify, and a suggestion you cannot trace is a claim you have to hope about. It is the difference described throughout the document intelligence reference, checkable beats impressive.
Both, in different roles, and keeping the roles straight is the point of the model. The fact is the unit of understanding, it is what the business acts on, an amount to pay, a quantity to price, a date to plan around. The document is the evidence, it is what makes the fact trustworthy and checkable. A well-built system lets you work at the level of facts while keeping every fact anchored to its document, so you get the usability of structured data and the accountability of the original paperwork at the same time. Losing either one breaks the model, facts without evidence cannot be trusted, evidence without extracted facts cannot be used.
Good document storage answers where is the file. The evidence model answers what does the file tell us about the job, and proves it. Storage leaves a person to open the PDF, read it and re-key what matters. The evidence model has already read it into facts, connected those facts to the job records they describe, and attached the source to each one. The same window described five different ways across the floor plan, the elevations, the schedule and the supplier quote becomes recognisable as one thing, backed by all five. That reconciliation, not the filing, is where the model earns its keep, and it is what the wider construction intelligence reference is built on.
Two things. Speed of trust, because a figure that shows its source is confirmed in seconds rather than taken on faith or re-derived. And a defensible record, because when a dispute or a query arises, the fact and the document behind it are already linked, so the evidence is in hand rather than buried in an inbox. In practice the gain shows up on the boring Tuesday when something is wrong and you need to see why, an evidence-backed system lets you follow any number straight to the page that produced it, which is exactly when a folder full of PDFs fails you.
05 / Terms
Glossary for this topic
Evidence model (treating a document as support for a fact rather than a file to store), provenance (a fact pointing back to the document and page it came from), fact (a discrete piece of meaning extracted from a document, the unit the business acts on), reconciliation (recognising the same fact described across several documents as one thing), source chip (an in-product link from a value to its evidence). The wider vocabulary lives in the construction glossary. From here the natural next article is document intelligence, which explains how documents become the facts this model relies on.
06 / Keep reading
Related knowledge, guides and features
Nothing hidden inside a PDF.
VIABUILD reads your documents into facts and keeps each one anchored to the page it came from, so every number Oryn shows you can be checked at a glance. Follow any figure straight to its evidence.