Evaluating Responses

Optimizing Your Copilot: The Power of Feedback

Feedback is the engine that drives InSkill’s intelligence. By evaluating the copilot’s responses, you help the system evolve from a general assistant into a precise technical expert. Your input is used to identify:

Knowledge Gaps: Areas where the documentation is missing or incomplete.
Incorrect Answers: Responses that are technically wrong or outdated.
Misunderstandings: Instances where the GPT misinterprets a question or provides a misleading answer.

The feedback system provides users with multiple ways to evaluate responses, helping improve accuracy, relevance, and overall performance. All users can rate responses with Thumbs Up 👍 or Thumbs Down 👎, while admin users with access to the Intelligence Portal can also flag 🚩responses and leave comments for further review.

How to Submit Feedback

Feedback is available immediately after every GPT interaction. There are two primary ways to provide input:

Method	Available On	When to Use
Thumbs Up	Portal + App	When the answer is accurate and helpful
Thumbs Down	Portal + App	When the answer is incorrect, misleading, or insufficient
Feedback Comment	Portal Only	Used exclusively with a Thumbs Down to explain the error

Feedback Options

All Users (App & Web)

👍 Thumbs Up – Positive Feedback

Select Thumbs Up if the response was:
✔ Relevant and accurate
✔ Clear and helpful
✔ Well-structured and complete

👎 Thumbs Down – Negative Feedback

Select Thumbs Down if the response was:
❌ Inaccurate or misleading
❌ Off-topic or irrelevant
❌ Unclear or incomplete

Intelligence Portal Users Only (Web)

🚩Flag – Escalating System-Level Issues

The Flag option is for cases where the response indicates a system-level issue rather than a lack of available content or incorrect configuration (e.g. roles/visibility/etc.). Flagging escalates the response for review by InSkill.

Thumbs down vs Flag: Thumbs-down ratings help identify potential areas for improvement but do not trigger immediate review by InSkill unless flagged.

💬 Comments – Adding Context to Flags

Users who flag a response can also add comments to provide additional details, such as:

Clarifications on why the response is problematic
Examples of expected output
Links to relevant source material that should have been referenced

How To: Thumbs Up & Thumbs Down

Web

App

What Happens Behind the Scenes?

When you submit a negative review, InSkill triggers an automated optimization workflow:

Intelligence Tab Entry: The problematic Question/Answer pair is automatically logged in the Issues tab under the Intelligence portal.
Review & Resolution: The author reviews the entry to determine the root cause:
- If it’s a Knowledge Gap: The author uploads a new resource (manual, video, or task) to fill the void.
- If it’s a System Error: The author marks the question with the FLAG icon in the Intelligence tab to signal a logic or processing issue for the engineering team.

Pro Tip: Providing a specific comment with your "Thumbs Down" is the single most effective way to ensure the same mistake isn't made twice.