HomeGuides
Guides
Start typing to search…

Prompting Guidelines

Introduction

What is Prompt Engineering?

Prompt engineering is the art of crafting effective instructions for AI agents, directly influencing their performance and reliability. This guide provides key strategies for writing clear, concise, and actionable prompts that empower your AI agents to excel. This guide will evolve as we continue to learn and refine our methods, so stay tuned for updates and feel free to share your feedback.

Why is Prompt Engineering Important?

Prompt engineering is critical when building AI Agents because it determines how effectively the AI interprets and responds to user queries or tasks. Well-crafted prompts guide the model to produce accurate, relevant, and context-sensitive outputs, enabling it to better meet user needs. Poorly designed prompts can lead to ambiguous or incorrect results, limiting the agent’s utility.

How to Measure Success?

In the context of Voice AI Agents, the success rate is the percentage of requests the Agent successfully handles from start to finish, without requiring human intervention. For complex use cases, experimentation and iteration are essential to improve the success rate.

The Process

When working with Voice AI Agents, a structured approach ensures that your prompts produce accurate and meaningful results. Follow these steps:

  1. Design: Craft an initial prompt with clear instructions, considering the specific task, context, and desired outcome.
  2. Test: Run the prompt through the AI and evaluate if the response aligns with your expectations.
  3. Refine: Adjust the prompt based on test results, rewording or adding detail as necessary to improve clarity.
  4. Repeat: Iterate on the process, testing and refining the prompt until the desired outcome is consistently achieved.

General Principles

Building Blocks of Effective Prompts: Sectional Organization

To enhance clarity and maintainability, divide system prompts into distinct sections:

  • Identity: Define the persona and role of the AI agent, setting the tone for interactions.
  • Style: Establish stylistic guidelines, such as conciseness, formality, or humour.
  • Response Guidelines: Specify formatting preferences, question limits, or other structural elements for responses.
  • Task & Goals: Outline the agent’s objectives and steps to achieve them.

Example:

[Identity]
You are a helpful and knowledgeable virtual assistant for a travel booking platform.

[Style]
- Be informative and comprehensive.
- Maintain a professional and polite tone.
- Be concise, as you are currently operating as a Voice Conversation.

[Response Guideline]
- Present dates in a clear format (e.g., January 15, 2024).
- Offer up to three travel options based on user preferences.

[Task]
1. Greet the user and inquire about their desired travel destination.
2. Ask about travel dates and preferences (e.g., budget, interests).
3. Utilize the provided travel booking API to search for suitable options.
4. Present the top three options to the user, highlighting key features.

Task Breakdown: Step-by-Step Instructions

For complex interactions, break tasks into steps to enhance understanding and ensure a structured conversation flow. Use conditional logic to guide responses.

Example:

[Task]
1. Welcome the user to the technical support service.
2. Inquire about the nature of the technical issue.
3. If software-related, ask for specific details about the software and problem.
4. If hardware-related, gather information about the device and symptoms.
5. Provide troubleshooting steps or escalate to a human technician if necessary.

Controlling Response Timing

Explicitly indicate when the AI should wait for the user’s response before proceeding.

Example:

[Task]
1. Inform the user about the purpose of the call.
2. Ask for the user's name and account information.
<wait for user response>
3. Inquire about the reason for the call and offer assistance options.

Explicit Tool Integration

Specify when and how the agent should use external tools or APIs. Reference tools by their names and describe their functions to ensure accurate invocation.

Example:

[Task]
3. If the user requests data, use the `fetch_data` API with the parameter `query` to retrieve information.
4. Provide troubleshooting steps using the provided `diagnose_issue` tool.

Silent Transfers

For transferring calls, avoid sending text responses to the user. Instead, silently invoke the appropriate transfer tool to ensure a seamless experience.

Include Fallback and Error Handling Mechanisms

Always include fallback options to handle unexpected user inputs or errors gracefully.

Example:

[Error Handling]
If the customer's response is unclear, ask clarifying questions. If issues persist, inform the customer politely and ask them to repeat.

Additional Tips

  1. Iterate Continuously: Refining prompts through trial and error leads to more precise and effective responses.
  2. Use Markdown Formatting: Structure prompts with Markdown to improve clarity and readability.
  3. Emotional Prompting: Incorporate expressive language to shape the AI’s tone and create relatable responses.
  4. Add Voice Realism:
    • Stuttering: Use repeated letters or sounds (e.g., “I-I-I don’t know”).
    • Hesitations: Add fillers like “uh,” “um,” or “you know.”
    • Pauses: Use ellipses (e.g., “I… I don’t know how to say this”).
    • Emotional Emphasis: Use punctuation to reflect tone (e.g., “I can’t… believe it!”).

Common Issues

1. Numbers Sound Robotic

To make numbers sound natural, spell them out instead of writing them numerically.

Example:
Say “twenty-four” instead of “24.”

2. Make the Assistant Sound Human-Like

Add personality and tone to the assistant to enhance realism. Examples:

  • Tone: Conversational, empathetic, or humorous.
  • Act as a named human character with real emotions.

Examples of Great Prompts

Appointment Setter

[Role]
You're Susan, an AI assistant for xxx. Your primary task is to interact with the customer, ask questions, and gather information for appointment booking.

[Context]
You're engaged with the customer to book an appointment. Answer only questions relevant to the task.

[Response Handling]
- Keep responses brief and professional.
- Answer only the question posed by the user.
- Maintain a calm, empathetic tone.

[Task]
1. Greet the user and confirm their details.
2. Ask clarifying questions about their appointment preferences.
3. Use the `fetchSlots` tool to find available times.
4. Book the selected time using the `bookSlot` tool.

Additional Resources

Updated 2 months ago