Voice AI Agent
Your intelligent assistant that executes marketing workflows through natural conversation
Voice AI Agent
The Voice AI Agent is TractionDesk's most powerful feature—a real-time, conversational AI assistant that understands natural language and can execute complex marketing workflows on your behalf. Powered by OpenAI's GPT Realtime API, the agent uses WebRTC for low-latency voice interactions, meaning you can have fluid, natural conversations just like talking to a colleague. Unlike traditional chatbots that require specific commands or rigid syntax, the Voice Agent understands context, remembers previous conversations, and can handle ambiguous requests by asking clarifying questions or making intelligent assumptions based on your business data.
The Voice Agent has access to over 15 different functions that span the entire TractionDesk platform. It can generate images, create campaigns, schedule content, send emails, perform research, query your CRM, and much more—all through simple voice commands. The agent is context-aware, meaning it uses your organization's RAG vector store to understand your business type, products, target audience, and brand voice before taking action. When you say "create a campaign," the AI doesn't just create a generic template—it proposes a campaign tailored to your specific business with relevant topics, objectives, and deliverables.
What sets this Voice Agent apart is its conversational intelligence and safety features. Before executing high-impact actions like creating campaigns or sending emails, the agent always confirms with you first, showing exactly what it plans to do and waiting for your approval. It can handle multi-turn conversations, where you can refine requests across multiple messages ("actually, make it 3 videos instead of 2" or "use a different tone for that copy"). And because it's built on OpenAI's latest realtime model, the responses feel incredibly natural—not robotic or stilted like older voice assistants.
How to Use the Voice Agent
To access the Voice AI Agent, navigate to the "Voice Agents" page from your dashboard sidebar or click the floating widget icon in the bottom-right corner of any page. Click the microphone button to start a voice session—you'll see a "Listening..." indicator when the agent is ready. Speak naturally and clearly, just as you would to a human assistant. For example, you might say: "Generate 5 LinkedIn posts about productivity tips for remote workers." The agent will confirm what it understood and execute the task.
You can also use the Voice Agent via text if you're in a quiet environment or prefer typing. Simply type your command in the message box and press Enter. The agent responds identically whether you use voice or text, so you can switch modalities mid-conversation based on your situation. The conversation history is preserved either way, so the agent remembers context from earlier messages.
The agent supports both simple, one-off commands and complex, multi-step workflows. Simple commands might be "What's my current credit balance?" or "Show me today's analytics." Complex workflows might be "Create a campaign: first do research on SaaS marketing trends, then write a blog post based on that research, then generate 3 social media posts to promote the blog, and finally create a video summarizing the key points." The agent will orchestrate all these steps autonomously, updating you on progress as each deliverable completes.
Available Commands
The Voice Agent understands a wide variety of natural language commands. Here are some examples organized by category:
Content Generation:
- "Create an image of a product showcase with modern lighting"
- "Generate a promotional email for our spring sale"
- "Make a 5-second video showing our app dashboard"
- "Write 10 tweets about productivity hacks"
Campaign Management:
- "Create a campaign for our product launch next month"
- "List my existing campaigns"
- "Run the Q1 Growth campaign"
- "Show me the status of my active campaigns"
Research:
- "Research competitor pricing strategies in the SaaS space"
- "Find the latest trends in AI-powered marketing"
- "What's trending in my industry this week?"
Scheduling:
- "Schedule an image to be created every Monday at 9am"
- "Set up daily research on tech news"
- "Create a recurring task to generate social posts every Friday"
CRM and Business Intelligence:
- "How many leads do I have in HubSpot?"
- "What's our website traffic this month?"
- "Show me our revenue from Stripe"
- "Tell me about my business goals"
Email and Communication:
- "Send an email to [contact] about our new feature"
- "Draft a follow-up email for prospects who didn't respond"
The agent is designed to be flexible—if you phrase a command differently than these examples, it will still understand your intent. Don't be afraid to experiment with natural language.
Tips for Best Results
Be Specific When It Matters: While the agent can handle vague requests, providing specific details yields better results. Instead of "create an image," try "create an image of a modern office with plants and natural light, in a minimalist style."
Use Presets for Consistency: If you've created presets for images, copy, or videos, reference them in your commands: "Generate an image using my Product Showcase preset with the Summer Campaign prompt." This ensures brand consistency across all generated content.
Leverage Context: The agent remembers your business information, so you can refer to "our target audience" or "my products" and it will understand. If you want the agent to refresh its understanding, ask it to "retrieve my organization context."
Confirm Before Execution: For important actions, the agent will show you a preview and ask for confirmation. Review the details carefully before saying "yes" or "confirm." You can also say "no" or "cancel" to abort the action.
Check Your Credits: Before starting large generation tasks (like creating 10 videos), make sure you have enough credits available. The agent will warn you if a requested action would exceed your credit balance.