The candidate assessment report is your primary tool for making data-driven hiring decisions. While manual review is often a key part of the process, Canditech’s AI Auto-Scoring provides an immediate layer of insight, offering a standardized evaluation of candidate responses for Open-Text, Video, and Coding questions.

Understanding how to read these results, explore the underlying logic, and apply your own judgment is essential for a seamless grading experience.

Before diving into the results, it’s helpful to know which "lens" the AI is using to evaluate your candidates. Canditech offers two distinct auto-scoring modes: Pre-built Agents and Custom Agents.

For Open-Text and Video Questions, pre-built agents evaluate specific competencies such as Conflict Resolution and Language Proficiency.

For Coding Questions, pre-built agents evaluate technical dimensions such as Time Complexity and Space Complexity.

Custom Agents allow you to define your own evaluation rubric and scoring criteria for Open-Text, Video, and Coding Questions.

For an in-depth guide into the AI Auto-Scoring options offered by Canditech, click here.

When you open a candidate’s assessment report, scroll to the specific question you wish to review. If AI Auto-Scoring was enabled, you will see a dedicated AI Results section below the candidate's answer.

For coding questions, the AI Results section appears alongside the coding test case results, allowing reviewers to evaluate both functional correctness and AI-generated assessments from a single location.

A single question can be evaluated by multiple AI models simultaneously (e.g., Conflict Resolution and Language Score, or Time Complexity and a Custom Agent). In these cases, the system distributes the weight equally: if two methods are used, each accounts for 50% of the question's total grade. If three are used, each contributes a third.

If you find a question with only the standard 5-star rating and no AI Results row, it simply means the question was not set for auto-scoring and requires your manual input.

For more information on reviewing candidates, click here.

Transparency is at the heart of our AI. By clicking on any of the scoring methods, you can expand it to see a detailed 1-5 star breakdown and a written justification. This rationale explains the "why" behind the grade. For example, if a candidate receives a lower score on a language model, the AI will specify whether it found significant grammatical errors or if the tone was simply inconsistent with professional standards. This allows you to understand the AI's perspective before you decide to accept or challenge it.

For coding questions, reviewers can similarly expand Time Complexity, Space Complexity, and Custom Agent results to review the reasoning behind each AI evaluation.

For those using the Custom Agent options, the interface becomes even more granular. Instead of just a general star rating, you will see a list of specific, binary rules that you defined for the role.

For coding questions, these rules may evaluate specific implementation requirements, coding standards, or technical constraints defined by your engineering team.

Next to each rule, a green checkmark (✅) or a red X (❌) will clearly indicate whether the candidate met that specific requirement. Just like the general models, each rule includes a detailed explanation of the rationale used to reach that conclusion. This format makes it incredibly easy to see exactly where a candidate excelled and where they fell short of your specific benchmarks.

While the AI provides a strong baseline, we believe the final hiring decision should always remain in human hands. If you disagree with an AI determination, overriding it is simple and intuitive.

For general star ratings: You can simply click on the number of stars (1–5) you believe is correct. This will immediately update the score to reflect your judgment.

For the Custom AI Agent: Because the score is built on specific rules, overrides are done at the rule level. If you disagree with the AI's "Pass" or "Fail" for a specific point, click the Thumbs Down icon. This will flip the result (e.g., changing a "Fail" to a "Pass") and automatically recalculate the total weighted score. You can always click the Thumbs Up icon to revert to the AI's original logic.

The star rating system converts your rating into a fixed percentage of the question’s total points. Each star corresponds to a percentage of the score, and that percentage is applied to the total points allotted for the question.

For example, if a question is worth 20 points:

This ensures consistent scoring across all manually reviewed questions, such as open-text and video responses.

For more information on reviewing candidates, click here.