AI Reasoning

The AI Reasoning feature enables the system to provide transparent, step-by-step explanations for its answers, rather than simply returning direct responses. When enabled, the AI reveals its thinking process, including how it analyses the question, retrieves and evaluates evidence, and arrives at conclusions. This feature leverages advanced reasoning agents such as ReAct (Reason-Action-Observation) and multi-hop retrieval, allowing the AI to break down complex queries, verify claims, and show confidence levels. AI Reasoning is especially valuable for users who require trust, auditability, or insight into how answers are generated. The feature can be toggled on or off according to user preference.

Use Case

Scenario

A user asks a complex question, such as comparing multiple products across several criteria, or verifying the accuracy of a claim found in uploaded documents. With AI Reasoning enabled, the system not only provides the answer but also displays the logical steps, sources, and verification results used in its analysis. This is ideal for research, compliance, education, or any context where transparency and evidence-based answers are critical.

Benefits

• Step-by-step reasoning is displayed for each answer.
• Confidence levels and verification summaries are included.
• Users can audit the process and see which documents or sources were used.
• Enhances trust and understanding of the AI’s outputs.
• Supports complex, multi-part, or ambiguous queries with structured logic.

Walkthrough

Enabling AI Reasoning

1. Go to Settings in the application.
2. Locate the AI Reasoning or Multi-Agent Reasoning toggle.
3. Switch the toggle ON to activate transparent reasoning for all answers.

Using AI Reasoning

• Ask any question as usual with AI Reasoning ON:
• The response will include a breakdown of the reasoning process, such as:
  • Initial interpretation of the question.
  • Step-by-step evidence gathering from uploaded documents.
  • Logical deductions or comparisons.
  • Verification of claims and confidence scoring.
  • Source citations for each step.
• For simple questions, the reasoning may be brief; for complex queries, the AI will show more detailed steps.

Reviewing Reasoning and Verification

• Each answer includes a Verification Summary showing how confident the AI is in its response and which evidence was used.
• Users can review the full reasoning chain and see which documents or sections contributed to the answer.
• For ambiguous or multi-part queries, the AI decomposes the question and tackles each part separately, showing its logic throughout.

Disabling AI Reasoning

• Return to Settings and toggle the feature OFF.
• The system will revert to providing direct answers without step-by-step explanations.

Note

AI Reasoning is designed for transparency and auditability. It is especially recommended for regulated environments, research, or any scenario where understanding the basis for an answer is important. The feature does not alter the underlying data or documents—only how the AI presents its findings.