Cited facts
Cited extraction is the evidence layer inside the Klarefi agent. It turns submitted files into facts your reviewer can verify: source quote, page, and status kept with the case.
PDF, scan, email
one packet, one case context
Per-fact citations
every value links to a quote and a page
Reviewer-first
disputed and missing facts surfaced separately
“This is not a separate product mode. It is the guarantee behind the agent: every extracted value stays attached to the source quote, the page, and the reviewer who can challenge it.”
How the agent turns uploaded files into facts your reviewer trusts.
- 01
Documents arrive through Hosted Intake or an enabled pilot API path.
- 02
Klarefi parses and classifies the packet against the workflow's fact requirements.
- 03
Required facts come out with source evidence: quote, page, bounding box.
- 04
Gaps and contradictions surface before a reviewer spends a cycle on them.
- 05
The final packet shows what is accepted, what is disputed, and what is still missing.
A reviewer should see the value and the source in the same glance. That is the entire product.
Where this evidence layer sits next to extraction tools.
- Alongside OCR
- Klarefi connects each page to the required fact and the review outcome that depends on it.
- Alongside IDP
- Klarefi pairs extraction with applicant follow-up so the case reaches review complete, not just parsed.
- Alongside your DMS
- Klarefi explains what the stored files prove and what is still missing, in the same workspace.
What document teams keep running into.
01
OCR delivers text and leaves the reading work on the reviewer's desk.
02
Extraction-only tools push the chase for missing evidence back to your team.
03
Manual review produces inconsistent notes and weak audit trails.
The verticals where this pays back fastest.
See it work on your own documents.
Run a pilot on your actual intake. Decide on measured accuracy and coverage, not on a demo.