Directors Console

Intent parsing and execution plan management

Directors Console

The Directors Console is the command center of AgencyOS, where human operators translate natural language intentions into executable marketing operations.

Overview

The Directors Console embodies the core philosophy of AgencyOS: human agency over all consequential actions. It provides an interface where AI assists in planning, but humans maintain final authority over execution.

Key Features

Natural Language Intent Input

Enter your marketing goals in plain English:

• "Create a LinkedIn post announcing our new product launch"
• "Generate social media content for our Q1 campaign"
• "Design banner ads for our holiday promotion"

The AI interprets your intent and generates a structured execution plan.

AI-Powered Plan Generation

The system analyzes your intent and:

1. Identifies required signals from your signal library
2. Generates execution steps with specific parameters
3. Estimates resources needed (time, API calls, costs)
4. Suggests optimizations based on past executions

Human Approval Workflow

Every generated plan requires explicit human approval:

• Review each step before execution
• Modify parameters if needed
• Add or remove steps to refine the plan
• Approve or reject the entire plan

This ensures you maintain control over all marketing operations.

Real-Time Execution Monitoring

Once approved, monitor execution in real-time:

• Step-by-step progress tracking
• Live status updates for each operation
• Artifact preview as they're generated
• Error notifications with suggested fixes

How It Works

1. Intent Parsing

When you enter an intent, the system:

// Example intent
"Create 5 LinkedIn posts about our new AI features"

// Parsed structure
{
  action: "create",
  content_type: "linkedin_post",
  quantity: 5,
  topic: "new AI features",
  context: {
    brand: "AgencyOS",
    tone: "professional",
    audience: "B2B marketers"
  }
}

2. Plan Generation

The AI generates an execution plan:

{
  steps: [
    {
      signal: "content-generator",
      action: "generate_linkedin_posts",
      parameters: {
        count: 5,
        topic: "AI features",
        tone: "professional"
      }
    },
    {
      signal: "content-reviewer",
      action: "review_content",
      parameters: {
        criteria: ["brand_voice", "accuracy", "engagement"]
      }
    }
  ],
  estimated_duration: "2-3 minutes",
  estimated_cost: "$0.15"
}

3. Human Review

You review the plan and can:

• Approve: Execute as-is
• Modify: Adjust parameters
• Reject: Cancel and start over

4. Execution

Once approved:

1. Each step executes sequentially
2. Progress updates in real-time
3. Artifacts are generated and stored
4. Final results are presented for review

Best Practices

Writing Effective Intents

Be Specific

• ❌ "Make some content"
• ✅ "Create 3 LinkedIn posts about our Q1 product updates"

Provide Context

• ❌ "Write a blog post"
• ✅ "Write a 1000-word blog post about AI in marketing, targeting CMOs"

Set Clear Goals

• ❌ "Do social media stuff"
• ✅ "Generate a week's worth of Twitter content promoting our webinar"

Reviewing Plans

Check Parameters

• Verify quantities, dates, and targets
• Ensure tone and style match your brand
• Confirm budget and resource allocation

Validate Steps

• Ensure logical flow of operations
• Check for missing steps
• Remove unnecessary operations

Consider Timing

• Review estimated duration
• Check if execution timing works for your schedule
• Plan for review time after generation

Integration with Other Components

Engine Room Connection

The Directors Console uses signals defined in the Engine Room:

• Signal Library: Available operations
• Templates: Pre-configured parameters
• Campaigns: Recurring execution patterns

Monitoring Dashboard

Execution results flow to the Monitoring Dashboard:

• Execution History: Past operations
• Artifact Gallery: Generated content
• Performance Metrics: Success rates and timing

Advanced Features

Template-Based Intents

Save frequently used intents as templates:

{
  name: "Weekly Social Media Batch",
  intent_template: "Create {count} {platform} posts about {topic}",
  default_parameters: {
    count: 5,
    platform: "LinkedIn",
    tone: "professional"
  }
}

Batch Operations

Execute multiple intents in sequence:

1. Define multiple intents
2. System generates combined plan
3. Review and approve all steps
4. Execute as a single batch

Conditional Execution

Set conditions for step execution:

• If/Then Logic: Execute step B only if step A succeeds
• Quality Gates: Proceed only if quality score > threshold
• Human Checkpoints: Pause for manual review at specific points

Troubleshooting

Intent Not Understood

If the AI can't parse your intent:

• Be more specific about what you want
• Use simpler language without jargon
• Break complex requests into smaller intents

Plan Generation Fails

If plan generation fails:

• Check signal availability in Engine Room
• Verify API keys for required services
• Review error messages for specific issues

Execution Errors

If execution fails:

• Check error logs in Monitoring Dashboard
• Verify input parameters are correct
• Ensure external services are accessible
• Retry with modified parameters

Security and Permissions

Access Control

• Role-based permissions control who can execute plans
• Approval workflows can require multiple approvers
• Audit logs track all executions

Data Privacy

• Sensitive data is encrypted at rest
• API keys are never exposed in logs
• Generated content is stored securely

Future Enhancements

Planned features for Directors Console:

• Voice input for hands-free operation
• Multi-language support for global teams
• Collaborative planning with team members
• AI learning from your approval patterns