Portal Walkthrough: Control 2.17 - Multi-Agent Orchestration Limits

Last Updated: January 2026 Portal: Copilot Studio, Power Automate Estimated Time: 1-2 hours per orchestration pattern

Prerequisites

Copilot Studio Environment Admin or Maker role
Power Automate license (for flow-based orchestration)
Multi-agent architecture documented
Delegation depth limits defined per zone
Circuit breaker thresholds established

Step-by-Step Configuration

Step 1: Document Agent Orchestration Architecture

Before configuring limits, document your multi-agent patterns:

Identify all agents that can invoke other agents
Map delegation chains (Agent A → Agent B → Agent C)
Determine maximum acceptable depth per zone:

Zone	Max Delegation Depth	Rationale
Zone 1	0 (no delegation)	Personal productivity, no chaining
Zone 2	2 levels	Team collaboration, limited chaining
Zone 3	3 levels	Enterprise, controlled orchestration

Step 2: Configure Agent-to-Agent Delegation

Open Copilot Studio
Select the orchestrating agent
Navigate to Actions or Plugins
If using agent-to-agent calls:
Document which agents can be called
Implement depth tracking in conversation context

Step 3: Implement Depth Tracking

For each orchestrating agent, implement depth tracking:

Conversation Variable: orchestration_depth
Initial Value: 0
Increment on each agent call
Check against max_depth before delegating

In Copilot Studio Topic:

Create a topic for delegation handling
Add condition node checking depth:
If orchestration_depth < max_allowed_depth: proceed
Else: return error or escalate to human

Step 4: Configure Circuit Breakers

Implement circuit breakers to prevent cascade failures:

Timeout Configuration:
Set maximum response time for delegated calls
Default: 30 seconds per agent call
Total chain timeout: 90 seconds
Retry Limits:
Maximum retries per failed call: 2
Backoff strategy: exponential
Failure Threshold:
After N failures, stop delegating
Default: 3 consecutive failures

Step 5: Configure HITL Checkpoints

For Zone 3 orchestrations, add human-in-the-loop checkpoints:

Identify decision points requiring human review
Configure approval topics:
Topic triggers on sensitive delegations
Pauses for human approval
Timeout with escalation

Step 6: Set Up Monitoring

Enable audit logging (Control 1.7)
Create alerts for:
Delegation depth exceeded attempts
Circuit breaker activations
HITL timeout escalations

Configuration by Governance Level

Setting	Baseline (Zone 1)	Recommended (Zone 2)	Regulated (Zone 3)
Max Delegation Depth	0 (none)	2 levels	3 levels
Circuit Breaker	Optional	Recommended	Required
Timeout per Call	60s	30s	30s
Total Chain Timeout	N/A	120s	90s
HITL Checkpoints	None	Optional	Required for sensitive
Monitoring	Basic	Enhanced	Real-time alerting

FSI Example Configuration

Orchestration Pattern: Client Service Request Handler

Primary Agent: Client-Service-Bot
Environment: FSI-Client-Services-Prod
Zone: 3 (Enterprise Managed)

Delegation Chain:
  Level 1: Client-Service-Bot
    → Delegates to: Account-Lookup-Agent
    → Delegates to: Transaction-History-Agent

  Level 2: Account-Lookup-Agent
    → Delegates to: KYC-Verification-Agent

  Level 3: KYC-Verification-Agent
    → No further delegation (leaf node)

Configuration:
  Max Depth: 3
  Total Timeout: 90 seconds
  Per-Call Timeout: 30 seconds

  Circuit Breaker:
    Failure Threshold: 3
    Reset Timeout: 60 seconds
    Half-Open Attempts: 1

  HITL Checkpoints:
    - Before KYC verification (customer data)
    - Before transaction execution (financial)
    - On confidence < 0.8

Monitoring:
  Alerts:
    - Depth exceeded: Immediate
    - Circuit open: Warning
    - HITL timeout: Escalation

Validation

After completing these steps, verify:

Delegation depth limits are enforced
Circuit breakers activate on failures
Timeouts are enforced correctly
HITL checkpoints pause for approval
Monitoring alerts are functioning
Audit logs capture orchestration events

Back to Control 2.17 | PowerShell Setup | Verification Testing | Troubleshooting