Verum resolves your policies, claims, brokers, and customers into one canonical graph — so every underwriting, distribution, and claims decision comes back as a plain-language answer, traced to the documents behind it.
Refer — POL-NVD93B (Commercial Property Shield · Property) shows a 3-yr loss ratio of 82% (red, >70%) and rate adequacy of just 88% — quoted €41k vs €47k technical — and the Property book is over its appetite cap. Send it to a senior underwriter before quoting.
acord-quote.xml · rating basis
loss-runs.xml · 3-yr experience
entities
8
dq flags
1
lineage depth
3
the demo
Four decisions, one graph.
The same canonical graph answers four everyday jobs — each a verdict you can act on, not a dashboard to read. Pick one in the trace above to watch it resolve, or open the real screen.
Three stages, one direction. Source-shaped records become a canonical graph you can interrogate — and every transformation stays on the record.
01
Ingest
Policies, claims, broker books, and ACORD filings land exactly as the source systems shaped them — XML, PDF, and everything in between.
02
Resolve
Duplicates and aliases collapse into canonical entities. Every link carries a confidence score; data-quality mismatches are flagged, not hidden.
03
Answer
Ask in plain English. The verdict is grounded only in that entity’s graph, its DQ findings, and its lineage — and it arrives with citations attached.
entity resolution
Three records. One company.
Brokers, policy admin, and claims each spell the insured a little differently. Verum folds them into a single canonical entity — and keeps the confidence and the evidence so a steward can always check the seam.
Behind each verdict is a PROV-O trail: the property value, the object snapshot it came from, and the source file it was derived from. Nothing is asserted that can’t be walked back to a document.
“Is Coverage C in force on POL-NVD93B?”
Yes — Coverage C is in force, sub-limit €250,000.
property_valuecoverage.inForce = true
object_snapshotCoverage C · POL-NVD93B
filepolicy.pdf · §Coverage C
prov·WAS_DERIVED_FROM · depth 4
governance
Built for people who get audited.
Roles that mean something
A Data Steward approves ingests and runs resolutions. An Analyst reads, asks, and inspects lineage. No one edits the record by accident.
ACORD as the backbone
The canonical model is aligned to ACORD object and link types, so what you answer with is what your market already speaks.
Quality kept in the open
DQ findings ride along with the entities they touch — surfaced on the node, counted in every answer’s footer, never quietly dropped.
Step into the graph.
Pick a seat. The demo runs on a real synthetic book — resolve entities, ask the graph, settle a claim.