Insurance
Unusual claims flagged, and every score explained
Flag unusual cases and explain every score.
Insurance and mutuals
Claims, policies, mutual insurers' health data: graal brings your data together on your infrastructure, trains models whose every score can be explained, and keeps a record of every version, every approval and every action.
What is at stake
Regulation (EU) 2022/2554 requires a register of information on arrangements with ICT service providers (Article 28), submitted to the ACPR every year.
Article 30 of DORA sets the minimum clauses of ICT contracts; for critical or important functions, they include an exit strategy.
Regulation (EU) 2024/1689 classifies as high-risk the AI intended for risk assessment and pricing in life and health insurance (Annex III, point 5(c)).
How graal responds
graal installs on your infrastructure; code is exportable and formats are open, which simplifies your exit strategy.
Every score comes with its reasons, and every model version is approved by a human before production.
Data, parameters, runs, approvals and agent actions: everything is logged, project by project.
Insurance
Flag unusual cases and explain every score.
Public sector
Query your internal documents with a model hosted on your premises, and answers that cite their sources.
Cross-industry
Move from Saagie, Hadoop or a SaaS by bringing your Python and Spark jobs along.
On your Kubernetes, on-premises or in your private cloud. For members' health data, with an HDS-certified hosting provider.
An 8-week pilot on a first use case, then an annual license per vCPU, unlimited users: a predictable cost, with no usage-based billing.
Compliance belongs to your organization. graal helps you trace and document: an identified vendor, an installation on your premises, exportable code, an audit of every action.
Yes, by installing graal with an HDS-certified hosting provider: the certification is held by the host.
The people the project grants that right to. Training sets can be pseudonymised, and personal data masking protects calls to language models.
A specialist walks you through a first use case and the traceability chain that comes with it.