Back to Documentation
Advanced Guide
12 min read

Guardian AI Query Guide

Getting the Most from AI-Powered Safety Analysis

Overview

Guardian AI is Safety Square's intelligent safety assistant. It operates in two modes:

  1. Job Step Analysis - Real-time hazard gap detection during job step creation
  2. Dashboard Insights - Natural language queries on safety data

This guide covers both modes and best practices.


Part 1: Job Step Analysis

When crews add job steps, Guardian AI analyzes the step and identifies gaps.

What Guardian Analyzes

InputAnalysis
Step descriptionWork type detection, inherent hazards
Selected hazardsCompleteness check against work type
Selected controlsAdequacy check against hazards
Project contextSite hazards, work type, environment

Guardian Output

  1. Safety Flags
    • Work types detected (hot work, electrical, etc.)
    • Permit requirements indicated
    • Special precautions flagged
  2. Missing Hazards
    • Hazards the crew may have overlooked
    • Based on step description and context
    • Ranked by relevance
  3. Suggested Controls
    • Controls to address identified hazards
    • Industry best practices
    • Site-specific recommendations

Example Analysis

Input:
  Step: "Weld bracket to column at 15-foot elevation"
  Hazards: ["Burns", "Fire"]
  Controls: ["Welding gloves", "Fire extinguisher nearby"]

Guardian Analysis:
  Safety Flags:
    - Hot work detected (permit required)
    - Elevated work detected (fall protection required)

  Missing Hazards:
    - Fall from height
    - Arc flash / UV exposure
    - Falling objects (to workers below)
    - Fume inhalation

  Suggested Controls:
    - Fall protection harness with 100% tie-off
    - Welding hood with appropriate shade
    - Barricade area below
    - Mechanical ventilation or respirator
    - Fire watch during and after work

Best Practices for Job Step Analysis

  1. Be Specific in Descriptions
    "Weld bracket at 15 feet" > "Do welding" - Details help Guardian understand context
  2. Include Location Information
    Height matters (fall hazards), area matters (confined space), environment matters (weather exposure)
  3. Don't Pre-Filter Hazards
    Select hazards you think apply, let Guardian catch what you missed. It's a safety net, not a replacement
  4. Review All Suggestions
    Don't auto-dismiss. Each suggestion has reasoning. Accept what applies, dismiss what doesn't

Part 2: Dashboard Insights

Safety Directors can query Guardian for patterns and insights across all safety data.

Accessing Insights

  1. Go to Dashboard
  2. Click "Guardian Insights" panel
  3. Type your question
  4. Review response

Effective Query Examples

Question TypeExample Query
Status Check"Which projects have active work right now?"
Hazard Analysis"What are the most common hazards across all projects?"
Compliance Check"Which crews haven't completed briefings today?"
Pattern Detection"Show me incidents involving falls this month"
Risk Assessment"Which projects have the most job steps without re-briefing?"
Comparison"Compare incident rates between projects"

Query Best Practices

  1. Be Specific About Time
    "today" / "this week" / "last 30 days" - Helps narrow results
  2. Name Projects When Relevant
    "Show me Unit 4 Turnaround activity" - Guardian knows your project names
  3. Ask Follow-Up Questions
    "Tell me more about that incident" / "Why is that project flagged?"
  4. Use Natural Language
    Don't try to code. Ask like you'd ask a colleague

Guardian Limitations

Guardian AI is powerful but has limits:

  • Not Real-Time Sensors: Guardian analyzes data, not physical conditions
  • Depends on Input Quality: Garbage in = garbage out
  • Suggestions, Not Requirements: Guardian advises, humans decide
  • Company Data Only: Can't access external databases or regulations

Ready to Transform Your Safety Execution?

See how Safety Square can help your team achieve session-first safety.