Mandacta · AI agent governance
Your AI agents act under mandate. Mandacta proves it.
Mandacta applies a discipline insurers already master to AI agents: delegation. Each agent receives a mandate — scope, authority limits, escalation thresholds — and reports to its principal, decision by decision.
The problem
AI agents already make decisions. Who authorised them, and how far?
AI agents are entering underwriting, claims handling and customer service. Their value is real, but one question still lacks a clear answer: who authorised this agent to do what, up to what amount — and who can prove it to the auditor, the regulator or the customer?
- Automated decisions without an explicit authorisation framework
- Technical logs that auditors and lawyers cannot read
- Human oversight that is claimed, rarely demonstrated
The principle
The mandate lifecycle
- 01
Mandate
Define for each agent its scope (products, covers, channels), its authority limits (amounts, rates, deductibles) and its escalation thresholds.
- 02
Control
Every action the agent proposes is checked against its mandate before execution. Outside the mandate, it is blocked or escalated.
- 03
Escalate
Beyond the threshold, the decision goes to an authorised employee, with all the context needed to decide.
- 04
Report
Chained, time-stamped log of every decision, periodic reporting to the principal, export for audit.
A language you already speak
From delegated underwriting to delegation to AI
| Classic delegation | With Mandacta |
|---|---|
| Binding authority agreement | Agent mandate, versioned and signed off |
| Underwriting limits | Authority limits enforced on every action |
| Referral to the insurer | Escalation to an authorised employee |
| Bordereaux and reporting | Decision log and automated reporting |
| Coverholder audit | Exportable audit trail, inspection-ready |
What Mandacta does
A register, a safeguard, a proof
Agent register
Inventory of AI agents in service, their mandates and their business owners.
Pre-execution control
Every action is systematically checked against the mandate in force.
Human escalation
Decision queues for out-of-mandate cases, with deadlines and permissions.
Decision log
Who decided what, on what basis, under which mandate — readable by non-engineers.
Reporting to the principal
Dashboards by agent, product and period; deviations and escalations highlighted.
Audit export
Evidence packs for internal audit, regulatory inspections and complaints.
Deployment
In your environment, not ours
Mandacta is a web application deployed in your environment, on premises or in your private cloud. Your policy, claims and customer data never leave your information system.
- Web application, nothing to install on workstations
- Deployed in your infrastructure
- Designed first for motor insurance
Regulation
From requirements to evidence
Human oversight, record-keeping, the right to human intervention in automated decisions: the EU AI Act and the GDPR (Article 22) set expectations. Mandacta produces the evidence that shows you meet them.
Mandacta
Pilot programme
Mandacta is under construction. We are opening a pilot programme to a small number of insurers, brokers and delegated authorities, to build the first release on their real cases.