Analyzing a property fund portfolio used to look like this
It looks like this now
Real estate does not lack information. It lacks a reliable way to reconstruct it.
An institutional portfolio is held through layers of entities, in several jurisdictions, under separate financing agreements. What it reports and what its own documents support are two different objects.
Reconciling them is manual, slow, and dependent on who does the work. Structural contradictions rarely appear inside one document. They emerge across them.
A contradiction that reaches the investment committee is a different problem from the same contradiction found in the record before it does. Every fact in a finding carries the document and clause it was read from, and every computed figure carries the facts behind it.
“One of the key issues faced by real estate companies is the unstructured labyrinth of real estate data.”
CBRE, How can AI bridge the gap for real estate investment managers? 2024
How a scan runs.
VDR, as delivered
- 01
Ingestion
Ingestion pipelineThe raw data room enters as delivered, classified and normalised.
- 02
Extraction
LLM-assistedFacts extracted under a fixed schema, with the source of every value recorded.
- 03
Structure reconstruction
DeterministicEntity, debt, temporal and income layers rebuilt in pure code, independent of the reported figures.
- 04
Coherence testing
DeterministicTests check whether the reconstructed structures can all be true at once.
- 05
Delta computation
DeterministicReconstructed values compared to reported ones, as a quantified delta.
- 06
Findings
DeterministicEach finding carries a complete evidence chain to specific clauses.
When two documents cannot both be true, ReActive shows you where.
One finding, traced to its sources.
- 09:12:07IngestionIngestion pipeline
- 09:38:20ExtractionLLM-assisted
- 09:41:02Structure reconstructionDeterministic
- 09:43:07Coherence testingDeterministic
- 09:44:19Delta computationDeterministic
- 09:45:02Findings identifiedDeterministic
Finding NTH-001 Surfaced structural exception
The planned exit window and the tenant break schedule fall in the same period
Model assumption
Disposal planned for H1 2027 at a 7.5% exit cap rate. The exit window is board confirmed; the cap rate is model sourced and is not board confirmed.
Assumptions Snapshot, Dec 2025 Board Meeting Minutes, Feb 2025
Documented reality
Break options across five tenancies fall inside that same window, exposing £2,846,924 of contracted rent to early exit.
Rent Roll, Dec 2024
Quantified delta
Evidence chain
- Rent roll, break scheduleNorthGate_Tenant_Schedule_Rent_Roll_Dec2024.pdf
- Board-confirmed exit windowBoard_Meeting_Minutes_Feb2025.pdf
- Model exit assumptionNorthGate_Financial_Model_Assumptions_Snapshot_Dec2025.pdf
Every displayed value traces back to the source document it was extracted from.
Open the demoSynthetic dataset representative of institutional real-estate operating conditions.
See a scan replayed.
The demo replays a recorded scan over a synthetic dataroom, from the documents to the findings.