IQ: AI engineer copilot
Learn how to use Quotient IQ for intelligent, feedback-aligned prompt optimization.
Welcome to IQ, an AI development copilot for writing, analyzing, and improving prompts through interactive sessions.
This documentation will guide you through the complete process of using IQ, from crafting your initial prompt to saving an optimized version.
📋 Workflow overview
Here’s a step-by-step guide on how to use IQ to improve your prompts:
- Create the initial user prompt
- Fill in prompt variables
- Run the generation and review the completion
- Provide feedback
- IQ suggests improvements
- Accept or reject the changes
- Run the new prompt and observe improvements
- Save the optimized prompt
📝 Step 1: Create the initial user prompt
The first step in using IQ is to create an initial user prompt. This prompt serves as a template that defines the instructions given to the AI. You can include placeholder variables that will later be replaced by actual input data.
For example:
🔄 Step 2: Fill in test prompt variables
Next, fill in test prompt variables. Variables allow you to insert dynamic content into your prompts, making it more flexible and reusable. This is especially useful when working with large datasets or varied inputs.
Example:
In this case, {{input}}
is the variable where you’ll enter a customer objection.
Run the prompt with the filled-in variables by clicking Run
.
✍️ Step 3: Run the generation and review the completion
Once you run the prompt, the model generates a completion based on the provided input. Review the completion to determine whether it aligns with your expectations.
If the completion is satisfactory, you can move to the next step. Should your response fall short of what you were expecting, you can provide feedback directly in the IQ interface as can be seen in the previous screenshot. Your feedback can highlight areas where the completion fell short or suggest how it could be improved. For example, if the response didn’t adequately address the customer’s concerns, you might leave a comment like:
IQ will take this feedback into account and suggest improvements to the prompt.
Note:
- IQ only considers negative feedback on the response. If the response is satisfactory, that’s great! However, there won’t be an option to provide additional feedback in such cases.
- From the moment you provide feedback, the response is automatically recorded as “Bad” in the
Run Log
and any linked datasets.
💡 Step 4: IQ suggests prompt improvements
Based on the feedback, IQ will automatically recommend changes to the prompt. These changes are designed to optimize the prompt and improve the quality of the AI-generated responses.
The suggested changes might include improving clarity, focusing more on addressing the customer’s specific objection, or modifying the structure of the prompt.
✅ Step 5: Accept or reject the suggested changes
After reviewing the suggested improvements, you have two options:
- Accept: If you like the changes IQ suggested, click Accept changes, and the prompt will be updated accordingly.
- Reject: If you prefer your original prompt, click Reject, and it will revert to the previous version.
This step allows you to maintain control over your prompt while benefiting from IQ’s intelligent suggestions.
🚀 Step 6: Run the updated prompt
Once you’ve accepted the changes, run the updated prompt again to observe how the AI’s responses have improved. IQ ensures a continuous cycle of improvement by iterating on feedback.
Keep testing, refining, and adjusting your prompts to reach the best possible performance.
💾 Step 7: Save the optimized prompt
Once you’re satisfied with the changes, save your optimized prompt. You can do this by clicking Save new version. This version is now stored for future use, ensuring that you have a well-optimized prompt ready for your task.
🔍 IQ best practices
Here are some best practices to follow when using IQ:
- Be specific in your feedback: The more specific your feedback, the better IQ can optimize your prompt.
- Use variables creatively: Variables allow you to create flexible prompts that adapt to various contexts, making them more reusable.
- Test iteratively: Keep running and refining your prompt based on real-world scenarios to achieve the most effective results.
For further assistance, feel free to reach out to our founders at contact@quotientai.co or check out more detailed documentation on PromptLab.