Cloud / AI Pilot Path
A controlled path to a Cloud/AI pilot
The recommended path is offline review, synthetic reference tenant, or read-only pilot before any production integration. Payment or interest alone cannot activate runtime.
Controlled request true
Live runtime false
Runtime enabled false
Auto remediation false
Pilot data classes
incident_historyMeasure cloud, SRE or AI runtime incident sequence, recurrence and closure signals.
Formats: csv, xlsx, json. Retention: pilot_retention.
Formats: csv, xlsx, json. Retention: pilot_retention.
cloud_metric_exportCompare weak cloud/runtime signals with alert timing, SLO pressure and operational decisions.
Formats: csv, json, log, txt. Retention: pilot_retention.
Formats: csv, json, log, txt. Retention: pilot_retention.
slo_summaryConnect SLO burn, error-budget pressure and governance thresholds to decision timing.
Formats: csv, xlsx, json. Retention: pilot_retention.
Formats: csv, xlsx, json. Retention: pilot_retention.
change_eventsQualify deployments, feature flags, configuration changes and rollback decisions without assuming causation.
Formats: csv, xlsx, json, log, txt. Retention: pilot_retention.
Formats: csv, xlsx, json, log, txt. Retention: pilot_retention.
ai_endpoint_metricsEvaluate AI endpoint latency, queue, error and capacity signals without prompt or customer content.
Formats: csv, json. Retention: pilot_retention.
Formats: csv, json. Retention: pilot_retention.
cost_exportModel aggregate FinOps exposure where the customer approves cost assumptions.
Formats: csv, xlsx, json. Retention: delete_after_review.
Formats: csv, xlsx, json. Retention: delete_after_review.
provider_statusKeep external provider status as context without using it as automatic proof or blame.
Formats: csv, json, txt. Retention: delete_after_review.
Formats: csv, json, txt. Retention: delete_after_review.
Acceptance checklist
cloud_ai_scope_and_sponsorpilot_scope
Acceptance: scope locked, human approval owner named, no automatic remediation accepted.
Acceptance: scope locked, human approval owner named, no automatic remediation accepted.
cloud_ai_data_requirementsdata_requirements
Acceptance: required classes approved, forbidden classes excluded, retention accepted.
Acceptance: required classes approved, forbidden classes excluded, retention accepted.
cloud_ai_security_and_data_boundarysecurity_boundary
Acceptance: DPA path defined, no credential material requested, audit trail ready, deletion rule accepted.
Acceptance: DPA path defined, no credential material requested, audit trail ready, deletion rule accepted.
cloud_ai_reference_tenantreference_environment
Acceptance: synthetic tenant present, app payloads present, reports present, no customer data.
Acceptance: synthetic tenant present, app payloads present, reports present, no customer data.
cloud_ai_report_packreporting
Acceptance: family catalog valid, review template valid, recurring reports valid, package ladder valid.
Acceptance: family catalog valid, review template valid, recurring reports valid, package ladder valid.
cloud_ai_public_research_pathpublic_research
Acceptance: noindex candidate, no customer data, no live write, no proprietary method disclosure.
Acceptance: noindex candidate, no customer data, no live write, no proprietary method disclosure.
cloud_ai_success_metricssuccess_metrics
Acceptance: metrics agreed, MODEL/ESTIMATE language accepted, no guaranteed ROI, human review cadence defined.
Acceptance: metrics agreed, MODEL/ESTIMATE language accepted, no guaranteed ROI, human review cadence defined.
Success metrics
Metric 01decision delay between signal and human action
Metric 02slo burn decision window
Metric 03inference degradation decision window
Metric 04change to rollback or scale decision time
Metric 05finops cost exposure with approved assumptions
Metric 06unknowns resolved before action
Metric 07decision defense completeness
Boundary
No sensitive credentialsNo cloud keys, cluster configuration files, provider admin credentials or sensitive credential material are requested before the gate.
No prompt contentAI endpoint metrics are aggregated operational signals; prompt and customer content are excluded by default.
No live writeRaven prepares decisions. A human approves. Automatic cloud remediation is not enabled.