Why Lumea
Traceability is the product.
Lumea runs on a provenance engine we call the Geltungskette: every output is bound to its source, authority, location, timestamp, and confidence level.
Anatomy of a Lumea output
Type
Planning - use class change
Document
Amtsblatt der Stadt Köln, 46. Jhg., G266/2025
Issuing office
Stadt Köln - Amt 60
Parcel
Mülheim, Flur 003, Flurstück 2785
Coordinates
50.96646°N · 7.00859°E
Discovered
14 May 2026 · 09:23 CET
Confidence
0.94 - Source-verified
AI-native, not AI bolted on
Built around AI from day one. Not retrofitted later.
Incumbent real-estate software was built for records, workflows, and dashboards. Adding an AI assistant on top does not make the system intelligent. It gives old software a new interface.
Lumea is AI-native: ingestion, extraction, interpretation, spatial matching, confidence scoring, provenance, and review workflows are part of the core architecture. The model is not a feature. It is one component in a system built to produce decision-grade intelligence.
That is the difference between using AI to decorate a product and building a product that could not exist without AI.
Domain-constrained intelligence
Source-bound outputs
Confidence over certainty
Human review when needed
The problem
Public does not mean usable.
In real estate, critical information is often technically public but practically unusable: scattered across portals, PDFs, notices, maps, registers, and local publication systems.
PropWatch starts with one concrete case: German planning and municipal publications. The legal standard is blunt: ordinary public notice can be enough, individual notification is not required, and owners can be expected to take precautions to stay informed.
Translation: the information may be legally public, but practically impossible to monitor at scale. One property is manageable. A portfolio across hundreds of municipalities is not - unless the process is automated, source-bound, and continuously updated.
OVG Rheinland-Pfalz, judgment of 10.06.2009 - 8 C 11319/08
The engine
Five links. No skipped steps.
Every Lumea output passes through the same provenance chain. If a link is missing, the output is rejected or held for review. Nothing reaches a customer without the full chain attached.
Geltungskette · v1.05 of 5 verified
01 · Source
Document
A PDF or notice published on a public municipal portal. Hashed on download for tamper evidence.
02 · Authority
Office
The office or body that issued it - planning department, city council, land registry, or responsible authority. Authority determines whether a document matters.
03 · Location
Parcel or asset
Bound to the relevant parcel, asset, or geography - not just approximate coordinates. The output touches a specific place.
04 · Time
Timestamp
Discovered-at and effective-from. Append-only. We can prove when we saw what.
05 · Relevance
Criticality
Each output is scored against the bound asset or location - regulatory shift, value impact, deadline urgency, operational relevance. Below the threshold, it logs but does not alert.
SHA-256 hashAuthority IDSpatial referenceISO-8601, CETScored 0-1
Responsible AI, European by design
Built on lawful sources. Designed for sensitive decisions.
Lumea’s principle is simple: AI systems used in strategic real-estate decisions must be lawful, traceable, privacy-conscious, and auditable from source to output. We build around European standards for data protection, source legality, hosting, and accountability - not as an afterthought, but as part of the architecture.
PropWatch applies this principle first to public planning, municipal, and spatial data. We process only sources we are entitled to process, bind every output to its provenance, and keep customer context separate from the public-source layer. Where personal or commercially sensitive data is involved, the system is designed for GDPR-aligned workflows, controlled access, and clear accountability.
For PropWatch, §5 UrhG matters because official works are excluded from copyright protection. Official notices, statutes, and planning documents can therefore be processed within the legal boundaries of public-source law. Case law also shows why monitoring matters: owners cannot always rely on individual notification and may be expected to take their own precautions to stay informed.
“Individual notification of affected owners is permissible, but not legally required.”
OVG Rheinland-Pfalz, judgment of 10.06.2009 - 8 C 11319/08
§5 UrhG · OVG Rheinland-Pfalz, judgment of 10.06.2009 - 8 C 11319/08
Why this matters
Trust has two sides: proof and completeness.
Generic scrapers and LLM-only systems can produce plausible answers. But in asset management, diligence, board discussions, or regulatory contexts, plausible is not enough. You need to know where an answer came from - and whether the system looked widely enough to find what mattered.
01 · Proof
Provenance
The Geltungskette turns extracted information into auditable intelligence: every output links back to the original source, the responsible authority, the relevant location, and the reasoning behind the confidence score.
02 · Completeness
Coverage
Critical developments rarely arrive as one clean document. They are scattered across portals, PDFs, notices, maps, registers, and local publication systems. A model can summarise what you give it. Lumea is built to find, index, bind, and score the needle before anyone asks the question.
PropWatch proves this architecture in one of the hardest real-estate data environments: public planning and municipal information across Germany. The hard work is not running a model on a PDF. The hard work is building a source index across Germany’s 11,054 municipalities, binding sources to locations, mapping responsible authorities, and scoring relevance so decision-makers see what actually matters.
“If we say it, we can prove it. If it matters, you will know.”Our Promise
Test the chain on an asset you know.
Send us an asset, parcel, or case you already understand. We will walk through the chain together - source by source, link by link - and you tell us where it breaks.
30 minutes · DE/EN · NDA on request