How it works
An agentic pipeline from evidence to explained verdict.
Four stages, running in seconds. The first three do the work a human investigator would do if they had the time; the fourth writes down why.
Capture
Guided evidence capture at the moment of return or first notice of loss — photos, video, documents, device and session signals.
- SDK and hosted capture flows for web, app and reverse-logistics stations
- Live coaching so the evidence is usable the first time
- Signed capture metadata: device, time, location, session lineage
Agentic review
Autonomous verification agents interrogate the evidence: image provenance, EXIF and tamper analysis, document forensics, policy and order matching.
- Specialist agents per evidence type, run in parallel
- Synthetic-media and re-render detection on images and video
- Document forensics across invoices, estimates and reports
Corroborate
Cross-checks against order history, carrier scans, repair estimates, network-level fraud signals and known-actor graphs.
- Order, payment and courier-scan reconciliation
- Identity, address and device linkage across the book
- Consortium-style repeat-actor signals, privacy preserving
Decide & explain
A verdict with a confidence score and a written, auditable rationale routed straight into your claims or returns workflow.
- Approve, refer or decline with per-signal weights
- Plain-language rationale an adjuster or regulator can read
- Webhook, queue or console hand-off with the full case file
The output
What a verdict actually contains.
A number on its own cannot be defended. Every Vericlaims decision returns the reasoning and the evidence alongside the outcome, so it can be reviewed by your team, explained to your customer and produced for your regulator without reconstruction.
Questions we get asked
The short answers.
How is this different from a fraud score?
A score ranks suspicion. Vericlaims verifies facts: whether the photo was taken now, whether the invoice was issued by that garage, whether the parcel that came back holds the item that left. The verdict is accompanied by the evidence that produced it, so it can be acted on and defended.
What do we need to integrate?
One API call to open a verification case, an evidence upload path (our capture SDK or your own), and a webhook to receive the verdict. Most teams run a shadow-mode pilot against historical cases first, then move to live decisioning.
Does it replace our investigators?
No. It removes the volume that never needed a human. Agents handle collection, checking and corroboration; your investigators get a prepared dossier and spend their time on the contested cases where judgement actually matters.
How do you handle personal data?
Evidence is processed for verification only, kept for a retention window you configure, and stored in the region you choose. Network signals are exchanged as derived indicators rather than raw personal data.
What happens when the agents are unsure?
The case is referred, not decided. You set the confidence thresholds per line of business, and every referral arrives with the specific signals that could not be resolved.
Which sectors are you built for?
Two today: returns and refunds in e-commerce and marketplaces, and claims in general insurance. Both run on the same verification core, tuned to the evidence and regulation of each market.
See it run on evidence that looks like yours.
A 30-minute walkthrough: we take a case type from your book, run the verification live, and show you the rationale it produces.