Claims / Agentic intake
The system drafts. A person promotes.
One queue of loss reports drafted from every non-manual channel — pasted text, email to your FNOL address, voicemail, text message, the guided wizard, bulk upload and partner systems — each waiting for a person.
“Drafts never become claims on their own — you promote or reject every one.” A draft that is wrong costs a correction, not an amended claim with a started clock.
The question this answers
What does a reviewer actually see?
The source, the fields the system extracted, the validation gaps it found, your organization’s fraud-rule results with their evidence, any duplicate hold, and the photographs. Promote creates the claim through the one claim door; Reject needs a reason; Re-run reads it again.
From report to reviewed claim
Read, check, promote
It Reads What Arrived
An email at 2 a.m., a pasted report, a voicemail, a text with photos. By morning it sits in the queue with the insured, policy, date of loss and description filled in.
Your Fraud Rules Run
Late reporting, prior claims, a loss close to policy inception or expiration, red-flag terms, possible duplicates — each rule kept, dropped and tuned by your organization.
Gaps Are Named
Validation gaps are shown on the item, so attention goes to what is missing or wrong rather than what read cleanly.
A Person Promotes It
On promote, the claim is created through the same door as a typed FNOL — same numbering, same checks, same assignment, same audit.
The boundary
What a flag can and cannot do
The value here is not the extraction. It is that an inaccurate draft cannot do damage before a person looks at it.
Fraud-rule Flags
Evidence for the person working it — flags never hold a claim
Flags After Promotion
Carried onto the claim’s SIU review
Duplicate Hold
Cleared or confirmed as a duplicate by a person
Unmatched Policy
Promoted unassigned, waiting in the Assignment Queue with the reason
Audit
Every promote, reject and duplicate decision is recorded
Entitlements
At least one intake add-on: Agentic FNOL, Guided Wizard, Voice FNOL, Text Message FNOL or the Integration Gateway
Where we are honest
What extraction cannot do
It reads reports well. It does not exercise judgment, and the design assumes it will sometimes be wrong.
It Does Not Create Claims
Only an item a person promotes becomes a claim. A clean text report is the one exception, and it runs through every standard check.
It Does Not Accuse
A fraud-rule hit is evidence, not an accusation, and it never holds a legitimate claim.
Poor Input Stays Poor
A faxed photocopy produces gaps. The honest outcome is more review, not a confident guess.
Read next
Send us your worst report
A scanned police report or a forwarded email chain. Watch what gets drafted, what gets flagged, and what a reviewer still has to fix.