Case studies · illustrative scenarios
How Intended would run on your shape.
Twelve shapes across SDLC, finance, security, healthcare, manufacturing, and government. Each is a constructed, illustrative scenario — not based on a specific customer deployment. The companies, figures, and outcomes are examples that show how the mechanism applies, not measured results.
fraud prevention
Every account freeze, limit change, and block checked against policy first, with step-up verification before a hard freeze. Audit trail exported for review.
Read →student data
FERPA-aware governance of AI advising, grading, and financial-aid agents. Each student-record read needs a policy decision first.
Read →deployment gates
Every production deploy waits on an authority decision from a gate step in the pipeline. A signed decision record per deploy.
Read →payment governance
Vendor payments in NetSuite gated through the SDK before the agent pays; the agent issues customer refunds in Stripe itself after an ALLOW.
Read →AI compliance
Adverse citizen decisions routed to a caseworker. Evidence export scores EU AI Act and NIST controls as satisfied, partial, manual, or not covered.
Read →patient data
PHI-touching agent actions governed by the healthcare policy pack. The EHR integration verifies the authority token before returning records.
Read →claims processing
Claims-agent payouts need verified coverage and an authority decision first. High-value claims escalate to an adjuster.
Read →contract automation
AI contract recommendations routed to the right attorney by value and clause type. Who approved what is on the record.
Read →supply chain
Procurement agents checked against vendor allowlists and budget before a PO exists. The agent creates approved POs in SAP with its own credentials.
Read →multi-tenant isolation
Tenant-scoped authority enforcement across customer-facing agent integrations. No cross-tenant signing surface.
Read →incident response
Pre-authorized P1 containment decisions for agents acting in CrowdStrike, Okta, and Vault. Every decision streamed to the SIEM by signed webhook.
Read →AI-native operations
Greenfield AI-native company whose agents ask Intended before every action. Every internal action is authority-bound.
Read →