ClearInsights Max: Best Practices

Effective use of AI and data can transform recruiting and performance management. This article shares best practices for asking clear, precise questions, training AI tools, and applying insights to drive better outcomes. Follow these strategies to make confident, data-driven decisions that empower your team.

In this article:
Recruiting Best Practices
Performance Best Practices
Additional Resources

Recruiting Best Practices

Effective question-asking is key when working with AI and data systems. Below are some essential tips for crafting clear, specific questions that enable AI to provide accurate responses. We’ll explore strategies like being explicit with details, choosing effective question formats, and training AI effectively. 

1. Be Specific and Clear: Use precise terms and avoid vague words like “bottleneck,” “efficient,” or “performing” that require interpretation. If the AI misunderstands, try rephrasing with more detail. If it mostly gets your answer but misses a few attributes, simply add to your question instead of rephrasing entirely. Familiarizing yourself with available attributes can help you ask better questions over time.
2. Define Dates Clearly: There are a variety of dates that can be used. If unspecified, AI will choose based on the selected attributes, past usage, and feedback. If you want a specific date type then specify in your question (e.g., “Apply Date”, “Entered Stage Date”, “Hired Date”, etc.). Without a defined time period, results will default to “all time.”
3. Specify Workflow Levels (Macro Stage vs Stage): When referring to these Workflows, be specific in which ones you want. Macro Stages are the high-level grouping of all your Stages.
4. Current vs. Historical Data: To see data for the current state of your Macro Stages or Stages, include “current” or “present” in your prompt. If unspecified, the system will default to historical data.
5. Use Exact Names: To ensure accurate results, always use the exact value names. At this time, AI relies on precise values to retrieve data, so even slight variations can cause errors or return no results. Being explicit helps the AI understand exactly what you need and avoids potential "hallucinations," where the system tries to interpret non-existent stages or sources. Here are some examples:
   • Macro Stage: If you want data for the "Interviewing" Macro Stage, use the full term “Interviewing,” not “Interview.”
   • Candidate Source: Use precise names, such as “ZipRecruiter” rather than “Zip Recruiter.”
6. Stage: If your stage includes “1st Phone Interview,” be specific and use the complete term, not just “Phone Interview.”
7. Top 1 or Bottom 1: When requesting a top or bottom result (e.g., top 1, bottom 1), the system shows only one value, even if multiple values match. For example, if you ask, “show me the recruiter with the fewest candidates in the interviewing macro stage”, and two recruiters each have the lowest count of 2 candidates in the interviewing macro stage, only one recruiter will be displayed. For more accuracy, request the top 3 or 5. However, if values are tied beyond your specified limit, only the first matching result will appear.
8. Easily Verify AI Output: Always verify the answer generated by AI by checking the Query Tokens. The output of AI is not 100% accurate, but can be trained to improve. The first step is always to verify both the attributes and any auto-generated formulas are correct.

Performance Best Practices

1. Be Specific and Clear: Currently, the AI is built for descriptive insights, not interpretation or reasoning. It won’t tell you why something is happening, so avoid broad or interpretive questions like “What are their strengths?” “Are there any patterns?” “What are the key predictors?”, or “What are the overall performance trends?” Instead, be specific about what you want to see.
   • For example:
   • “What are the average review scores by department?”
   • “Which goals were marked complete in the last 30 days?”
   • “How many employees received a score below 3 in their most recent review?”
   • The conversational UI works like building blocks—you don’t have to get everything right in one question. Start with a simple query, then refine by adding or clarifying details. This iterative approach helps you narrow in on the insight you need without starting over.
   • Over time, getting familiar with the available attributes will help you ask sharper, more effective questions.
2. Define Dates Clearly: There are many types of dates in your data, and the AI doesn’t assume which one you mean. If you don’t specify a date field (like “Review Cycle Close Date”, “Goal Due Date”, or “Survey Response Date”), the system will guess based on context—using selected attributes, past usage, or feedback.
   • To get accurate results, name the exact date field you want. Also, if you don’t specify a time frame (e.g., “last quarter” or “past 30 days”), the AI will default to all-time data, which may not match what you’re looking for.
3. Use Exact Names: For accurate results, always use the exact value names. The AI currently relies on precise matches, so even small differences can lead to errors or no results. Being specific helps prevent confusion or "hallucinations," where the system interprets data that doesn't exist.
   • Example: If a review is titled “2025 Q1 Goal Review”, you must use that full title. Typing just “Q1 Goal Review” may not return the correct data.
4. Top 1 or Bottom 1: When you request the top or bottom result (e.g., “top 1” or “bottom 1”), the system returns only one value—even if multiple values are tied. For example, if two employees in Sales have the lowest review score of 2, only one will be shown.
   • To capture ties or get a broader view, ask for the top 3 or top 5 instead. However, if more than your requested number of shares have the same value, only the first matching result will be displayed.
   • Tip: The system does not know whether higher or lower numbers are better. If your scoring scale runs from 1 (best) to 5 (worst), asking for “best scores” may return employees with a score of 5. To avoid confusion, use terms like “highest” or “lowest” for clarity.

Additional Resources

ClearInsights Max: Recruiting Getting Started
ClearInsights Max: Performance Getting Started