Control Plane · AI
AI & Atlas Settings
The dials that govern Atlas. Confidence, sensitivity, cadence, and what the model learns from your operation.
Atlas modelReasoning v2.0 · COO persona
Auto-execution≥ 95% confidence · impact ≤ $25k
Last calibration9 days ago · 87 outcomes scored
Confidence threshold
Show recommendation70%
Below this, Atlas keeps observations private.
Requires approval85%
Above this, single-approver path is offered.
Auto-execute eligible95%
Combined with impact cap.
Alert sensitivity
P0 (Critical)
AlwaysBusiness hours
P1 (High)
AlwaysBusiness hoursDigest
P2 (Watch)
AlwaysBusiness hoursDigest
P3 (Info)
AlwaysDigestOff
Recommendation cadence
New recommendationsEvery 15 minutes
Re-rank existingHourly
Stale auto-archive48 hours · with summary
Quiet hours22:00–06:00 local · P0 overrides
Executive digest
CadenceDaily · 06:30 local
ChannelsIn-app · Email
AudienceCEO · COO · CIO · VPs
FormatBrief Template v1.1 · 5 cards
Learning preferences
Learn from approvals
Learn from rejections (with reason)
Learn from executive comments
Share anonymised outcomes with industry benchmark
Model refreshWeekly · Sundays 02:00 ET
Guardrails
Prohibited actionsWorkforce reduction without HR approval
Compliance scopeHIPAA · SOX · GDPR
PII redactionOn · evidence is masked in briefs
Hallucination checkCitation required for every claim
Outcome tracking
Every recommendation Atlas issues is graded against the actual operational outcome.
Recommendations · 30d
412
Approved
318 · 77%
Auto-executed
89 · 22%
Outcome accuracy
91%