No-code fraud rules and thresholds — including AI-generated rules from a plain-language description, always reviewed before activation.
Every investigator decision — confirmed fraud or false positive — becomes training data, so the next score is sharper. Retrain in one click on your own team’s decisions.
Community detection, bridge analysis that compares every provider, doctor and member with its peers, and time travel to replay how a fraud network grew day by day.
See where fraud concentrates — drill down from country to province, provider and physician, each measured fairly against its own volume.
A plain-language case brief written by AI for every claim, with each risk score explained by the factors behind it and a human-approved recommendation.
Risk ranking and anomaly detection.
A complete case file from first alert to recovery: interview log, recovery tracking, secure evidence attachments, a visual timeline and a one-click case report — plus optional bulk AI review when your team chooses it.
Executive dashboards and reporting.
Ask questions in plain language, by text or voice, and get an instant answer — no SQL, no dashboards, no analyst in the loop.
Specialized agents for rule generation, case decisions, executive summaries, bridge analysis and platform administration — human approval by default, automation by choice.
Connects to your claims system on Oracle, SQL Server, MySQL, PostgreSQL or CSV files, transferring only changed records — no changes to your core systems.
Flag known bad-actor providers, doctors and members the moment they appear in an alert, a case or the fraud graph — managed by your own team, with full audit history and no impact on unrelated claims.
Every alert is transparent, auditable and supported by risk scoring, graph analytics and AI-driven investigation. Our graph engine uncovers hidden provider networks, detects communities of related claims and flags cross-community bridge providers that connect otherwise separate fraud rings.
Our AI case agent proposes a decision and rationale for every case — but never acts alone: every recommendation requires human review and approval before it becomes a record, keeping investigators in full control and every decision auditable.
Executives can also ask the platform directly: our conversational AI assistant answers plain-language questions about claims, providers and alerts — by voice or text, in any of our 8 supported languages — running on an isolated, read-only connection so it can never modify your data.
Healthcare data is sensitive. InsurGraphAI lets each organization decide which AI it uses — and guarantees that personal data never leaves the building.
Use a leading cloud AI service, your organization’s own AI server, or a local model running inside your data center — or switch AI features off entirely. Changed from the admin screen, no code required.
For air-gapped and public-sector environments, the local model option runs fully on-premises without a GPU requirement.
Names and national ID numbers are masked before any request leaves your organization; the answer is restored on your own server. Analysts inside the organization still see full details.
Route AI traffic through your own corporate proxy and compare it with the built-in send log — byte for byte, your security team can confirm that no personal data is sent out.
Built for insurers and public institutions that need control, traceability and full ownership of their data. Security details and answers to your security questionnaire are available on request.
Deployed on your own servers or in the cloud — your choice. Automated installation on Windows Server or Ubuntu.
In an on-premises deployment, claims and personal data never leave your network. Only masked text reaches an external AI service — and only if you enable one.
All traffic between users and the platform runs over HTTPS (TLS).
User passwords are stored only as salted hashes; AI service keys are stored encrypted and shown masked on screen.
Active Directory sign-in, role-based permissions, automatic sign-out after inactivity and one active session per user.
Who did what, and when: sign-ins, rule and user changes, AI settings and scheduled runs — searchable, exportable and protected against deletion.
Rule engine, model training and network refresh run automatically every night at the time you choose.
English, Turkish, Dutch, German, French, Spanish, Arabic and Chinese — including AI-generated reports, voice questions and read-aloud answers.
Deterministic analysis runs first; AI is called only when it adds value, and unchanged results are never generated twice.
Faster Reviews
SIU Productivity
Leakage Reduction
Detection Increase
Talk to the people who build and bring InsurGraphAI to insurers.
Email: info@insurgraph.ai
LinkedIn: https://www.linkedin.com/company/insurgraphai/