Universal Auto Triage

AI agent · Workflow / back-office agent · IT & helpdesk by Torq

Agentic AI engine for ruthless threat prioritization and noise reduction.

Universal Auto Triage is an agentic AI engine for security operations that prioritizes threats and filters noise by learning from a team's past decisions. It helps SOC teams reduce duplicate alerts by 90% and process telemetry at machine speed.

Agent specs

The full spec card (context, pricing per token, modalities and more) on AntSpace. Sign up free to see the full spec card.

Languages
English

Category: AI Agents

Data sources
SIEM, EDR, Cloud, Identity, and application security tools
Core Engine
Agentic AI
Learning Type
Proprietary historical decision learning
Integration Count
400+ security tools
Input Formats
AWS Finding Format, Azure Alert Schema, GCP SCC Finding
Management Interface
Socrates Builder (Natural Language)
  • 90% reduction in duplicate alerts
  • Triages alerts in as little as 56 seconds
  • Learns from your SOC's historical decisions
  • Natural language rule creation via Socrates Builder

Overview

Universal Auto Triage is the agentic AI engine within the Torq AI SOC Platform designed to prioritize security threats and eliminate alert noise for enterprise security teams. It acts as the primary ingestion point for the SOC, automatically normalizing telemetry from over 400 security tools and deduplicating events to reduce fatigue. The engine uses proprietary technology to learn from a team's historical decisions, continuously refining its accuracy to deliver verdicts on actual risks at machine speed. Security analysts can manage the engine using natural language through the Socrates Builder interface to create custom triage rules and guidance. By grounding decisions in a continuously updated context model of the environment, it ensures that verdicts are accurate and auditable.

Triage and Prioritization

  • Agentic Prioritization: The agentic AI engine prioritizes security threats and separates noise from actual risk at machine speed.
  • Continuous Learning: Proprietary technology learns and remembers how a specific SOC works, refining accuracy based on past decisions and judgment.
  • Telemetry Normalization: The engine ingests and normalizes telemetry from across the security stack, correlating and deduplicating events to reduce alert fatigue.
  • Risk-Based Verdicts: It analyzes risk context and threat intelligence to deliver verdicts that separate false positives from actual risk.

Intelligence and Governance

  • Natural Language Rules: Users can create guidance and rules for the engine using plain English descriptions via the Socrates Builder interface.
  • Contextual Grounding: Decisions are grounded in a continuously updated context model of the environment, including identities, assets, and relationships.
  • Transparent Reasoning: The engine provides transparent audit logs of its reasoning and actions for human oversight.
  • Human-in-the-Loop Control: Manual override options ensure that security analysts remain in control of the final decisions.

What it does

  • Telemetry ingestion and normalization
  • Event correlation and deduplication
  • Risk context analysis
  • Threat prioritization
  • False positive suppression
  • Alert verdict generation
  • Triage rule creation via natural language

Published results

  • 90% duplicate alert reduction
  • 75% faster alert processing
  • Triaged alerts in 56 seconds for a global pharma giant

Use cases

  • Security analysts use the engine to automatically filter out false positives and focus on high-priority threats.
  • SOC managers deploy the engine to normalize telemetry across multi-cloud environments like AWS, Azure, and GCP.
  • Detection engineers use natural language to create and refine triage rules based on new threat intelligence.
  • Incident responders rely on the engine's transparent reasoning to understand why an alert was prioritized or suppressed.

Specifications

Integrations
AWS Security Hub, Azure Sentinel, GCP Security Command Center, CrowdStrike, Wiz, Microsoft Sentinel, SentinelOne, ServiceNow, Abnormal AI, and 400+ tools
Data sources
SIEM, EDR, Cloud, Identity, and application security tools
Human control
Operates autonomously with transparent audit logs and manual override options; guidance and rules are reviewed before enabling.
Models
A hybrid of proprietary Torq models and leading frontier models
Core Engine
Agentic AI
Learning Type
Proprietary historical decision learning
Integration Count
400+ security tools
Input Formats
AWS Finding Format, Azure Alert Schema, GCP SCC Finding
Management Interface
Socrates Builder (Natural Language)
Governance Features
Transparent audit logs and manual override

How it works

  1. Telemetry Ingestion: The engine ingests telemetry from across the security stack, including SIEM, EDR, and cloud platforms.
  2. Correlation and Deduplication: Events are normalized, correlated, and deduplicated to remove noise and identify related incidents.
  3. Agentic Analysis: The agentic AI analyzes risk context and threat intelligence to deliver a verdict on the alert's legitimacy.
  4. Prioritization and Routing: Verified threats are prioritized and routed to the appropriate response teams or automated workflows.

FAQ

How does the engine learn from my team?

Universal Auto Triage uses proprietary technology to learn and remember how your specific SOC works, improving its accuracy and speed over time based on your team's past decisions.

Can I create my own triage rules?

You can use Socrates Builder to attach a playbook or describe your requirements in plain English to draft new entries, which you can review before enabling.

Is the AI's decision-making transparent?

The engine provides transparent audit logs of its reasoning and offers a manual override option so analysts always maintain final control.

What tools does it integrate with?

It integrates with over 400 tools, including major cloud platforms like AWS, Azure, and GCP, as well as SIEM and EDR solutions.

Added by AntSpace from public sources (torq.io).

Related

More AI agents by Torq