Research Project: This is a free AI research project. No warranties, SLAs, or company associations. Learn more

The OODA Loop

UnderstandingHow AI Does RMM

The OODA loop is a military decision framework: ObserveOrientDecideAct. Every agent runs this loop continuously, learning from every outcome.

The Continuous Loop

Click each phase to learn how the brain processes information and takes action. This loop runs every 10 seconds, ensuring rapid response to changing conditions.

Loop cycles every 10 seconds with live telemetry

Observe Everything

TelemetryEventsMetrics

The edge agent streams real-time telemetry from every endpoint—CPU, memory, disk, processes, services, network, security events.

  • Lightweight Go agent on every endpoint
  • Sub-second telemetry streaming
  • osquery for deep system introspection
  • No polling—continuous observation
Live Simulation

See It In Action

Watch the brain process a real-world scenario: High CPU on a production server.

Signal: High CPU Alert

ACME-DC01 • CPU at 98% for 15 minutes

Observe
Orient
Decide
Act
Risk-Tiered Autonomy

Confidence Gates Action

Not all actions are equal. The brain assesses risk and requires appropriate approval before executing.

Read Only

Gather information, no changes

System inventory
Log collection
Performance metrics
Auto-execute

Low Risk

Minor changes, easily reversible

Restart service
Clear temp files
Update config
Auto-execute

Medium Risk

Significant changes, may need rollback

Kill process
Install patch
Firewall rule
Confidence > 85% or Admin

High Risk

Major changes, business impact possible

Reboot server
Isolate endpoint
Rollback update
Admin approval required

Critical

Dangerous actions, irreversible

Wipe data
Decommission
Emergency lockdown
Dual-owner approval
Evolutionary Learning

Every Outcome Teaches

The brain doesn't just act—it learns. Every decision and outcome is recorded, building institutional knowledge that improves over time.

Step 1

Action Executed

Brain takes action based on OODA decision

Step 2

Outcome Observed

Was the problem fixed? Did anything break?

Step 3

Result Recorded

Decision + outcome stored in memory

Step 4

Confidence Updated

Success increases confidence; failure decreases

Step 5

Pattern Learned

Similar situations get better recommendations

Loop repeats, getting smarter with each cycle
Vector Memory

Semantic Search Across All Decisions

Every decision becomes a 1024-dimensional vector. Similar problems cluster together, enabling instant pattern discovery.

Embedding Every Decision

Using OpenAI's text-embedding-3-large model, every decision, outcome, and pattern is converted to a 1024-dimensional vector and stored in pgvector with HNSW indexing.

embedding = model(decision_context)

Semantic Similarity Search

When a new signal arrives, the brain finds similar past situations instantly—not by keyword matching, but by understanding the meaning.

similar = brain.search(signal, top_k=5)

Confidence From History

Past outcomes directly inform current confidence. If similar actions succeeded before, confidence increases. If they failed, the brain adjusts.

confidence = aggregate(similar_outcomes)

Cross-Fleet Pattern Discovery

With tenant consent, anonymized patterns from across the fleet improve everyone's experience. When a fix works for one MSP, similar MSPs benefit from that knowledge—without exposing sensitive data.

Traditional RMM vs. Agentic RMM

See the fundamental difference in approach

Traditional RMM

  • ×Alert fires → Human reviews → Human decides → Human acts
  • ×Knowledge lives in technician heads
  • ×Same problems recur with same manual fixes
  • ×Reactive: Wait for problems to happen
  • ×Evidence: Scattered across tickets and emails

Agentic RMM

  • Signal detected → AI reasons → AI decides → AI acts (with gates)
  • Knowledge in evolutionary brain, accessible to all
  • Patterns learned, confidence grows, outcomes improve
  • Proactive: Predict and prevent before impact
  • Evidence: Cryptographic chains for every action

Ready to See It In Action?

Watch the OODA loop process real telemetry, make decisions, and take action—all in a live demo.