Best Practices for Building AI Metrics and AI Process Prompts in Oversai

In Oversai, creating powerful and effective AI metrics / AI Processes begins with well-structured, clear prompts. The quality of these prompts directly impacts how accurately the AI can evaluate, analyze, and provide actionable insights. To achieve the best results from your AI-driven processes, follow these best practices for building your AI metrics:

1. Be Specific and Clear

Ensure that your prompts are precise. The AI needs detailed instructions to understand exactly what it's assessing. For example, instead of asking the AI to "analyze the conversation," specify it to "evaluate the agent's empathy during the conversation."

2. Use Boolean Metrics (Yes/No) for Precision

For optimal performance, create AI metrics with a single yes/no question. This allows the AI to produce clear and direct answers, enhancing the accuracy of the results. If you need to ask more questions, it's a good practice to create multiple AI metrics rather than combining several questions into one. This will ensure the platform remains highly precise and provides more valuable insights, as each AI metric targets one specific element.

3. Incorporate Contextual Prompts

Provide the AI with enough context to make accurate assessments. Contextual clues—such as tone, behavior, and situational factors—are essential for evaluating customer interactions effectively. The more context you provide, the better the AI will understand the nuances of the interaction.

4. Focus on Outcomes, Not Just Actions

Metrics should go beyond just identifying whether an action was performed. They should evaluate the quality and effectiveness of that action. Your prompts should guide the AI to analyze how well the action met the customer’s needs or resolved the issue at hand.

5. Iterate and Refine

AI models improve over time, but it’s crucial to regularly refine your prompts to keep them aligned with your evolving business goals. Review feedback and assess the AI’s performance frequently, making adjustments when necessary.


Additional Tips:

  • Avoid Ambiguity: Make sure your prompts are clear and leave no room for misinterpretation. A well-defined question ensures that the AI understands exactly what it's analyzing.

  • Keep it Simple: Use short, focused questions. Simplicity ensures that the AI stays concentrated on evaluating specific criteria within the interaction.

  • Test and Tweak: Once your AI metrics are set up, test them with real data to see if they’re accurately capturing the desired outcomes. Adjust as needed to ensure precision.

  • One Question per AI Metric: For each AI metric, it is recommended to include only one yes/no question so the AI can effectively resolve it. If you need to ask more questions, you can create additional AI metrics. This approach ensures the platform is as precise as possible and provides greater value.


Example Metrics

Below are revised examples, each containing only one question per AI metric to ensure clarity and focus:

Title: Emotional Intelligence

  • Type: Boolean (Yes/No)

  • Guideline: Did the agent demonstrate empathy by acknowledging and validating the customer's feelings or frustrations?

Title: Troubleshooting

  • Type: Boolean (Yes/No)

  • Guideline: Did the agent accurately diagnose the root cause of the issue?


By following these best practices and focusing on a single yes/no question per AI metric, you can create highly precise and impactful metrics within Oversai. This will lead to better insights and enhanced performance for your customer support teams while making the most out of AI-driven processes.



Still need help?

Contact us

Best practices