Autonomy Levels
Control how much ReductrAI does automatically. From detection-only to full auto-remediation.
Human-in-the-loop by default. ReductrAI proposes actions. You approve. Trust is built over time.
The Four Levels
AI monitors and alerts. Human does everything else.
AI Does
- Monitor telemetry 24/7
- Detect anomalies
- Send alerts
Human Does
- Investigate root cause
- Decide on remediation
- Execute fix
Available on: All tiers
AI investigates and finds root cause. Human decides and executes.
AI Does
- Everything in L1
- Trace dependencies
- Correlate events
- Identify root cause
Human Does
- Review AI's findings
- Decide on remediation
- Execute fix
Available on: All tiers
AI investigates AND proposes remediation. Human approves with one click.
AI Does
- Everything in L2
- Propose remediation options
- Assess risk levels
- Calculate blast radius
- Execute after approval
Human Does
- Review proposal
- Click Approve or Reject
- No manual execution
Available on: All tiers (default)
AI executes known patterns automatically. Human is notified after.
AI Does
- Everything in L3
- Execute pre-approved runbooks
- Auto-remediate known patterns
- Notify after action
Human Does
- Define allowed runbooks
- Review post-incident
- Adjust policies
Available on: Business, Enterprise, BYOK only
Comparison
| L1 Detect | L2 Investigate | L3 Approve | L4 Auto | |
|---|---|---|---|---|
| Anomaly detection | ✓ | ✓ | ✓ | ✓ |
| Root cause analysis | - | ✓ | ✓ | ✓ |
| Remediation proposals | - | - | ✓ | ✓ |
| One-click execution | - | - | ✓ | ✓ |
| Auto-execution | - | - | - | ✓ |
| Human approval required | N/A | N/A | Yes | No* |
* L4 auto-executes only for pre-approved runbooks
Configuration
Autonomy level is configured in the ReductrAI dashboard. Navigate to Settings → Autonomy to change your level.
Valid levels: L1, L2, L3 (default), L4
Why L3 is the Default
- AI does 95% of the work - investigation, correlation, proposal
- Human retains 100% control - nothing executes without approval
- Compliance-friendly - full audit trail of who approved what
- Trust builds over time - start with L3, graduate to L4
Graduating to L4
L4 auto-remediation is earned, not given. The recommended path:
- Start at L3 (approve mode)
- Build a track record - AI proposes, you approve
- Identify patterns that AI consistently gets right
- Create runbooks for those patterns
- Enable L4 for specific runbooks only
- Expand L4 as trust grows
L4 is not "set and forget." It's a partnership. AI handles the known. Humans handle the novel.