Governance · 6 min read
Fact, inference and action. Designing trustworthy property intelligence
Trust depends on separating what is observed, what is assessed and what is recommended.
Property intelligence only earns trust if its users can tell what kind of statement they are looking at. Three kinds matter, and they should never be blurred together.
A fact is a verified, source-backed observation with a source and a timestamp. An EPC rating retrieved from the official register on a known date is a fact. Its provenance is part of its value.
An inference is a transparent assessment with logic, uncertainty and confidence. An AVM estimate, a placeability score or a resilience flag is an inference. It may be excellent, but it is not a fact, and it should be labelled with the reasoning and confidence behind it.
An action is a task with an accountable owner and an override path. It is the point where intelligence becomes operational, and where audit matters most.
No automated recommendation should be presented as an unquestionable fact. Making the distinction visible in the interface is a governance control, not a design detail. When users can see the difference, they can calibrate their trust, challenge the right things and defend the outcome later.
This separation is what allows automation to grow safely. Facts can flow quickly. Inferences can inform routing. Actions stay owned by people, with every completion, exception and override recorded.