Trust & compliance
Verification you can defend.
Vericlaims sits at the point where money moves and a customer is told yes or no. That means the decision has to survive a complaint, an audit and a regulator — so the platform is built to produce evidence of its own reasoning, not just an outcome.
Explainable by default
Every verdict ships with the evidence, the checks that ran and the reasoning. No unexplained scores in front of a regulator.
Human in the loop
Agents recommend; your team decides on edge cases. Escalation thresholds are yours to set, per line of business.
Data minimisation
Evidence is processed for verification purposes only, with configurable retention windows and regional data residency.
Built for audit
Immutable decision logs, versioned rule and model lineage, and exportable case files for disputes and litigation.
Security
How evidence is handled.
Evidence attached to a return or a claim is some of the most sensitive material your customers will ever send you. It is processed for verification and nothing else.
Encryption
Evidence and case data encrypted in transit and at rest, with key management separated from application access.
Regional processing
Region-pinned endpoints and storage so evidence stays inside the jurisdiction you operate under.
Least privilege
Scoped API credentials, role-based console access, and access logging on every case record.
Retention you set
Configurable retention windows per case type, with verifiable deletion at the end of the window.
Immutable decision logs
Every verdict, override and access event written to an append-only log with model and rule lineage.
Tested continuously
Independent penetration testing and a responsible-disclosure channel for anything found in the open.
How we govern the agents.
Automated decisioning about consumers is regulated, and it should be. These are the rules the platform holds itself to, independent of what any individual customer configures.
- Agents recommend; declines and adverse outcomes stay under human authority unless you explicitly configure otherwise
- Every automated decision carries the checks, findings and reasoning that produced it
- Model and rule versions are recorded per decision, so a past verdict can be reconstructed exactly
- Outputs are monitored for differential error rates across customer groups, and reported to you
- Evidence submitted for verification is not used to train general-purpose models
Diligence pack
Security questionnaires, architecture detail, data-flow diagrams, subprocessor list, current certification and attestation status, penetration-test summaries and our DPA are available to prospective customers under NDA.
We would rather tell you exactly where we are than imply more than we can evidence — which is, after all, the whole point of the product.
Security issues can be reported to the same address. We will acknowledge within one business day and will not pursue good-faith researchers.
Bring your security team to the first call.
Most verification conversations stall on data handling. We would rather have that conversation at the start than three months in.