AI Validation for Sentinel Queries: Smarter KQL with Uncoder AI

How It Works

This Uncoder AI feature automatically analyzes and validates detection queries written for Microsoft Sentinel using Kusto Query Language (KQL). In this example, the input is a multi-condition search query designed to identify domain names linked to the SmokeLoader campaign (CERT-UA references shown).

The left panel shows the detection logic:

search (@”dipLombar.by” or @”dubelomber.ru” or @”iloveua.in” … )

The query uses literal string matching to detect specific threat domains.

The right panel displays AI-generated validation output, where Uncoder AI dissects the query into its syntactic and semantic components:

• Correct use of KQL syntax (search , @ for literals, or operators).
  
• Performance implications (e.g., large number of OR conditions, no wildcard usage).
  
• Schema correspondence advice for better data alignment.
  

Suggestions for maintainability (e.g., using in operator or joining from a lookup table).

Explore Uncoder AI

Why It’s Innovative

Security engineers often work under pressure and lack the time or context to deeply review the technical and performance aspects of every query. Traditionally, detection queries are:

• Written ad hoc without optimization.
  
• Rarely documented or performance-tuned.
  

Uncoder AI solves this by:

• Parsing the query structure with LLMs trained on KQL and detection engineering best practices.
  
• Providing clear, actionable suggestions — not just rule correctness, but better ways to query based on data volume and use case.
  

This elevates Uncoder AI beyond code generation — it becomes an expert assistant embedded in the detection pipeline.

Operational Value

For SOC teams and detection engineers, the benefits are immediate:

• Reduced trial-and-error: Validation ensures logic runs as expected before deployment.
  
• Higher performance: Optimized syntax improves efficiency at scale.
  
• Cross-skill enablement: Even junior analysts gain expert-level insights into KQL usage.
  
• Faster tuning: AI advice accelerates detection refinement cycles across environments.
  

In essence, Uncoder AI doesn’t just write queries — it thinks with you, validates in real time, and enables precision detection engineering in platforms like Microsoft Sentinel.

Explore Uncoder AI