Show the artifact,
not just the outcome story
Exemples illustratifs de la façon dont KLA gouverne des workflows d'agents réglementés et des preuves qu'il produit pour les auditeurs.
Governed workflow exemplars
Each example is an illustrative scenario — not a customer deployment — structured around the evaluation questions that appear in real buying committees.
Treasury approval workflow
Banking · illustrative scenario
What the team needs
A bank InfoSec team needs hard stops on high-value transfer requests and named approvers for exceptions.
How KLA intercepts
KLA intercepts transfers above threshold, attaches account context, and routes the decision to treasury leadership in Slack.
Evidence it exports
- Policy hit: `payments.high_value_requires_human`
- Approver identity and approval timestamp
- Signed downstream transfer execution record
{
"action": "wire_transfer.create",
"decision": "requires_human_review",
"policy": "payments.high_value_requires_human",
"approver": "[email protected]",
"lineage_hash": "sha256:4f3a...a18e"
}Clinical support escalation
Healthcare · illustrative scenario
What the team needs
Clinical governance needs human review whenever a copilot moves from summarisation into recommendation territory.
How KLA intercepts
KLA blocks unapproved recommendation delivery, preserves the patient-safe context window, and routes the case to a clinician.
Evidence it exports
- Reviewer assignment and decision outcome
- Model and prompt version used in the recommendation path
- Execution lineage showing no unauthorised action reached the patient channel
{
"action": "care_plan.recommend",
"decision": "blocked",
"reason": "clinical_recommendations_require_review",
"review_queue": "clinical-oversight",
"lineage_hash": "sha256:91cd...77be"
}Claims settlement control path
Insurance · illustrative scenario
What the team needs
Claims leadership wants AI triage and drafting, but legal needs review for settlement recommendations above internal authority limits.
How KLA intercepts
KLA routes recommendations by authority band, captures the reviewer path, and records exactly which claim context and policy rules shaped the outcome.
Evidence it exports
- Authority threshold and rule evaluation result
- Reviewer identity and settlement disposition
- Signed lineage chain for the claim action and final outbound message
{
"action": "claim.settlement_recommendation",
"decision": "approved",
"authority_band": "review_required",
"reviewer": "[email protected]",
"lineage_hash": "sha256:2c6e...bb90"
}