Guide to Defining AI Agent Objectives & Prompts
A well-configured AI agent needs a clear objective and structured prompts to ensure effective and meaningful interactions. In Awaz.ai, you can:
✔ Set a clear objective that defines the agent’s primary task (e.g., booking calls, answering FAQs, lead generation).
✔ Customize prompts to control conversation flow and guide the AI’s responses.
✔ Test and refine prompts to optimize the agent’s accuracy and engagement.
This guide walks you through the Agent Objectives and Prompt Settings, ensuring your AI works exactly as intended.
Step 1: Define the Purpose of Your Agent
Before setting up an AI agent, it’s essential to determine its main role. Ask yourself:
What is this AI agent supposed to do?
Will it handle customer inquiries, schedule appointments, or provide automated support?
Should the AI initiate calls or wait for user input?
Once the purpose is clear, you can define it in the Agent Objective field within the Settings tab.
🔹 Where to Set the Agent Objective
Log into Awaz.ai and navigate to the Agents section.
Select an existing agent or create a new one.
Click on the "Edit" and on the Settings tab, locate the Agent’s Objective field.
Enter a short but clear description of what the AI should accomplish.
💡 Tip: Be specific and focus on the AI’s main role. Example:
✅ Good Objective: "Schedule appointments for demo calls and confirm bookings via SMS."
❌ Too Vague: "Help customers."
Step 2: Configuring Prompts & Conversation Flow
Once the objective is set, you’ll need to configure how the AI interacts with users. This is done under the Prompt tab.
🔹 Setting Up AI Prompts
Go to the Prompt tab in the agent settings.
Choose who speaks first:
User Initiates – AI waits for the user to start.
AI Initiates (Dynamic Greeting) – AI begins with a flexible introduction.
AI Initiates (Defined Greeting) – AI starts with a specific, scripted message.
Set a custom greeting to personalize interactions.
Use Freestyle Prompting to add:
Background Info – Key details the AI should know.
Instructions – Rules to keep the AI’s responses aligned with its role.
Scripts – If needed, pre-defined structured conversations.
💡 Tip: The Company info should include relevant details like the company name, services, and key talking points.
✅ Example: "XYZ Consulting is a global advisory firm specializing in digital transformation and AI-driven automation solutions. We help enterprises optimize workflows, enhance customer engagement, and streamline operations using cutting-edge AI technology."
🔹 Training AI to Recognize Unique Names & Pronunciations
One common issue in AI prompting is how it pronounces unique names, brands, or locations. The AI may misread or misinterpret words unless it is trained properly.
✔ Example Issue: A company name “Hotel45” may be pronounced as "Hotel Four Five" instead of "Hotel Forty-Five".
✔ Solution: In the Freestyle Prompting section, specify pronunciation by writing it phonetically or providing a guiding instruction.
✅ Example Fix in Freestyle Prompting:
"When mentioning 'Hotel45,' pronounce it as 'Hotel Forty-Five' instead of 'Hotel Four Five.' Maintain the correct pronunciation across all responses."
💡 Tip: Use Boosted Keywords in Voice Settings for critical brand names, industry-specific terms, and common words the AI needs to transcribe accurately.
📌 Looking to optimize AI responses? Check out our Guide to Configuring AI Voice Settings.
Step 3: Save & Test the Configuration
After setting up the objective and prompts, it’s crucial to test your AI agent.
🔹 How to Test Your AI Agent
Click Save Changes after setting up objectives and prompts.
Use the Train your Agent panel on the right side to run sample interactions.
Review AI responses and adjust prompts if needed.
Conduct real-world trials before deploying the agent live.
💡 Tip: If the AI isn’t responding as expected, refine the instructions or freestyle prompts for better accuracy.
Need Help?
We’ve got plenty of resources to assist you!
📺 Tutorial Videos: Watch step-by-step guides
🔑 Login / Sign-up: Access Awaz.ai
📖 Help Center: Visit our knowledge base
📧 Support Email: [email protected]