Agent Builder Wizard

Design your first voice agent with guided prompts for purpose, personality, workflows, and success metrics.

Agent Builder Wizard

This interactive tutorial guides you through designing a complete voice agent from concept to specification. By the end, you'll have a comprehensive agent design document ready for implementation.

Time to complete: 20-30 minutes

How It Works

The wizard walks you through the Five Pillars of Agent Design with guided questions, industry templates, and real-time validation. Your answers are compiled into a downloadable design document.

The Five Pillars

  1. Purpose & Scope — Define what problem your agent solves
  2. Personality & Voice — Establish tone, style, and brand alignment
  3. Conversation Workflows — Map primary conversation paths
  4. Data & Tools — Identify required APIs and capabilities
  5. Success Metrics — Define how you'll measure performance

Tutorial Interface

Note:

Interactive Tutorial Coming Soon

This tutorial is currently in development. The wizard will include:

  • Step-by-step guided questions
  • Industry-specific templates (hospitality, e-commerce, support)
  • Real-time validation and suggestions
  • Side-by-side examples from Otto and Cypress
  • Downloadable design document (JSON + Markdown)
  • Option to start implementation directly

For now, use the Agent Design Guide to manually complete your agent design.

Preview: Tutorial Flow

Step 1: Choose Your Industry

Select an industry template to pre-populate common patterns:

  • Hospitality — Hotel bookings, restaurant reservations, concierge services
  • E-Commerce — Product sales, recommendations, shopping assistance
  • Customer Support — Ticket resolution, FAQ automation, troubleshooting
  • Healthcare — Appointment scheduling, patient intake, insurance verification
  • Real Estate — Property inquiries, showing scheduling, lead qualification
  • Custom — Start from scratch for unique use cases

Step 2: Define Purpose

Answer guided questions:

  • What specific problem does this agent solve?
  • Who will interact with it? (customers, employees, partners)
  • What does a successful interaction look like?
  • What should the agent NOT do?

Example Output:

{
  "purpose": "Enable hotel guests to search available rooms and complete bookings without phone calls or forms",
  "target_users": ["website visitors", "returning guests", "travel planners"],
  "success_criteria": ["booking completed", "availability confirmed", "guest information collected"],
  "out_of_scope": ["cancellations", "complaint handling", "loyalty program management"]
}

Step 3: Establish Personality

Define tone, style, and brand voice:

  • Formal or casual?
  • Energetic or calm?
  • Concise or detailed?
  • Empathetic or neutral?

Example Output:

{
  "tone": "warm, attentive, professional",
  "style": "like a luxury hotel concierge — anticipate needs, confirm details",
  "boundaries": "doesn't handle cancellations or complaints (escalates to human)",
  "sample_phrases": [
    "I'd be delighted to help you find the perfect room",
    "Let me check availability for you"
  ]
}

Step 4: Map Workflows

Build conversation flows with visual flowchart builder:

  1. Greeting & Intent Detection
  2. Information Gathering
  3. Tool Execution (search, book, confirm)
  4. Confirmation & Next Steps
  5. Edge Cases & Escalation

Example Output:

{
  "primary_flow": [
    "Greet warmly and detect booking intent",
    "Gather check-in/out dates and guest count",
    "Call search_units tool",
    "Display available rooms with photos",
    "Confirm selection and create reservation",
    "Show booking confirmation"
  ],
  "edge_cases": {
    "no_availability": "Suggest alternative dates or room types",
    "user_changes_mind": "Gracefully restart search",
    "complaint": "Empathize and escalate to human support"
  }
}

Step 5: Plan Tools & Data

Identify required integrations:

  • What real-time data does the agent need? (inventory, pricing, availability)
  • What actions must it perform? (create booking, send email, update CRM)
  • What APIs provide this data?
  • What authentication is required?

Example Output:

{
  "data_requirements": [
    "Real-time room availability by dates",
    "Room details (photos, amenities, pricing)",
    "Guest reservation management"
  ],
  "tools": [
    {
      "name": "search_units",
      "purpose": "Query available rooms by check-in/out dates",
      "api": "https://your-hotel.com/api/units/search",
      "auth": "Bearer token"
    },
    {
      "name": "create_reservation",
      "purpose": "Book a room for the guest",
      "api": "https://your-hotel.com/api/reservations",
      "auth": "Bearer token"
    }
  ]
}

Step 6: Define Success Metrics

Set measurable goals and operational limits:

  • Conversation completion rate target
  • Average session duration
  • Escalation rate threshold
  • Concurrent session limits
  • Budget constraints

Example Output:

{
  "metrics": {
    "completion_rate": "60%+",
    "avg_session_duration": "less than 5 minutes",
    "escalation_rate": "less than 10%",
    "customer_satisfaction": "4.5/5 or higher"
  },
  "limits": {
    "max_concurrent_sessions": 20,
    "max_session_duration": "15 minutes",
    "daily_token_cap": 300000
  }
}

Export Options

When you complete the wizard, you can export:

  1. Design Document (Markdown) — Human-readable specification for stakeholders
  2. Configuration JSON — Machine-readable format for direct import
  3. Implementation Checklist — Step-by-step tasks for developers
  4. Agent Prompt Starter — Pre-filled system prompt template based on your design

Using Your Design Document

After completing the wizard, proceed to implementation:

Tips for Success

Start Narrow

Your first agent should do one thing excellently. You can expand scope after validation.

Use Templates

Industry templates provide proven patterns. Customize from there rather than starting blank.

Think Real Scenarios

Map workflows based on actual customer interactions, not hypothetical ideal cases.

Plan for Failure

Define edge cases and escalation paths before implementation.

Need Help?

If you get stuck during the wizard:

  • Click the "?" icon next to any question for examples and guidance
  • Use "Load Example" to see how Otto or Cypress answered each section
  • Save your progress and return later
  • Reach out to support@strategicmachines.ai for design consultation

Ready to design your agent? Start the Agent Builder Wizard (Coming Soon)

For now, manually work through the Agent Design Guide.