Creating a New Prompt
Step 1: Navigate to the Prompts Section
Log into your Kakiyo dashboard
Click on the Prompts tab in the left navigation menu
Click the Add Prompt button in the top right corner
Step 2: Configure Basic Prompt Settings
Enter a Name for your prompt (e.g., "Job Offers," "Demo Request")
Select an AI Model from the dropdown menu (e.g., GPT-4.1, Claude 3.7 Sonnet)
For Type, select "Messages" (currently the only available type)
Click Continue to proceed to the prompt editor
Step 3: Write Your Prompt Content
The prompt editor is divided into two main tabs:
Context Tab
This is where you provide instructions for the entire conversation:
Start with a Role section:
# Role
[Describe who the AI is representing and in what context]Define the Mission:
β# Mission
[Explain what the AI should try to achieve through the conversation]Add Instructions for the AI:
β# Instructions
[List specific guidelines for message style, length, tone, language, etc.]Include Special Cases handling:
β# Special Cases
[Provide instructions for handling objections, questions, or situations]
First Message Tab
Here you specifically instruct the AI on crafting the initial outreach message:
# Task
[Briefly describe what the first message should accomplish]
# Constraints
[List important limitations or requirements for the first message]
Make sure to include the required variables {{productDescription}}
and {{prospectDescription}}
somewhere in your either the first message prompt or the context prompt.
DO NOT FILL THOSE VARIABLES β the system will handle it for each prospect.
Step 4: Configure Variables
On the right side of the prompt editor, you'll see the Variables panel with three sections:
Required Variables:
prospectDescription
andproductDescription
are pre-defined and mandatoryEnsure these are used somewhere in your prompt content
Note: These are automatically filled by the system - you don't need to provide values for them
Optional Variables:
These include
agentName
,prospectName
,companyName
Also automatically filled by the system
You can use them in your prompt, but they're not required
Custom Variables:
You can add custom variables like
goal
,goalLink
,fallback
These make your prompt more flexible and reusable
Important: Unlike required and optional variables, you will need to manually fill in values for these custom variables when setting up your campaign
To add a new custom variable:
Type its name in the "New variable name" field
Click the "+" button to add it
Use it in your prompt with double curly braces:
{{variableName}}
Testing Your Prompt
Step 1: Access the Test Interface
Click the Test Prompt button in the top right corner of the prompt editor
This opens the test conversation interface
Step 2: Configure Test Settings
Select a Product from the dropdown menu
Fill in values for all variables in your prompt
Click Start Test to begin the conversation simulation
Step 3: Simulate a Conversation
The AI will generate the first message based on your prompt
You can respond as if you were the prospect
The AI will reply according to your prompt instructions
Continue the conversation to test different scenarios:
Test positive responses to see how the AI guides toward your goal
Test objections to see how the AI handles resistance
Test irrelevant questions to assess how the AI redirects the conversation
Step 4: Refine Your Prompt
Based on the test results, return to the prompt editor to make improvements
Adjust instructions, wording, or variables as needed
Test again until you're satisfied with the AI's performance
Click Save Changes to save your refined prompt
Using Premade Prompts
Kakiyo offers a library of pre-configured prompts that you can use as starting points:
When creating a new prompt, look for template options
Select a template that matches your campaign goals
Customize the template to fit your specific needs
Test and refine as needed
Best Practices for Effective Prompts
Be Specific: Vague instructions lead to inconsistent performance
Test Thoroughly: Simulate various prospect responses
Iterate: Continuously improve your prompts based on test results
Keep it Natural: Avoid overly sales-focused language that sounds robotic
Focus on Value: Guide the AI to highlight benefits rather than features
Include Fallbacks: Always provide alternative paths for uninterested prospects
Troubleshooting
If the AI generates inappropriate responses, check your constraints section
If messages are too long, specify word count limits more clearly
If conversations feel unnatural, add more guidance on conversational tone
If the AI isn't following your instructions, ensure they're clearly structured with numbered points