How Narwal Used Snowflake Cortex AI to Transform Unstructured Safety Notes into Proactive Workforce Intelligence

Summary

A leading American manufacturer relied on manual tracking and unstructured safety notes to monitor workplace safety, creating a 5-to-7-day lag in identifying critical behavioral trends and coaching opportunities. While traditional safety metrics captured lagging indicators like Total Recordable Incident Rate, they failed to uncover the underlying themes and fluid conversation patterns necessary for proactive risk mitigation.

To bridge this gap, Narwal built a  Safety Conversation Intelligence Assistant powered by Snowflake Cortex AI and advanced enterprise semantic modeling. Safety and operations teams can now automatically extract themes, behavior patterns, and coaching opportunities from unstructured text. Safety notes that were once passive records are now transformed into searchable, measurable intelligence in less than 30 seconds, keeping all sensitive compliance and workforce data securely inside Snowflake.

Customer Challenge

The manufacturer generated a high volume of safety logs, near-miss reports, and operational observations across its facilities. While these unstructured narratives contained vital safety insights, they were functionally locked behind manual workflows. Because safety managers had to manually review thousands of text entries, identifying systemic risks introduced severe operational friction.

• Delayed Risk Identification: Critical behavioral risks and safety themes went unnoticed for days, preventing safety teams from intervening before hazards escalated into recordable incidents.

• Subjective Coaching Opportunities: Frontline supervisors lacked a standardized, data-driven framework to identify specific behavioral trends that required targeted safety coaching.

• Static Reporting Limitations: Existing EHS dashboards could only track structured data such as dates and department codes, completely missing the rich operational context buried inside conversational notes.

The safety and management teams needed answers to critical, fluid questions on a daily basis:

• What are the recurring behavioral risk themes being reported across our facilities this week?
• Which specific operational workflows show a statistically significant increase in safety coaching opportunities?
• Are there shifting behavior patterns in our safety notes that correlate with recent shift changes or new line introductions?

Narwal’s Solution

Narwal designed a unified, Snowflake-native architecture powered by Snowflake Cortex AI that extracts, classifies, and analyzes unstructured safety text in real time, built around core generative AI capabilities.

Automated Conversation & Note Enrichment

Safety logs and text narratives are automatically processed using AI_CLASSIFY and specialized analytical workflows. The assistant turns unstructured text into enriched metadata, categorizing safety notes by hazard types, risk levels, and behavioral categories without requiring manual human tagging.

Trusted Enterprise Safety Accuracy

Consistency was achieved through a safety-specific Semantic View and Cortex Search. This standardized layer ensures the assistant accurately interprets complex manufacturing and EHS terminology, aligning text retrieval with the manufacturer’s specific risk management and OSHA-aligned compliance metrics.

Unified Text Retrieval & Structured Analysis

Safety intelligence requires connecting what people say with hard operational data. The solution combines advanced text retrieval with structured analysis over enriched metadata, allowing users to query both the qualitative context of safety notes and quantitative safety trends simultaneously.

Context-Aware Analytical Deep Dives

Powered by Snowflake Agents and Cortex Analyst, the assistant maintains deep context across multi-turn conversations. Safety professionals can ask follow-up questions, drill down into specific risk themes, and explore safety trends without restating complex parameters.

Business Outcomes

• Turned Text into Measurable Intelligence: Unstructured safety notes and field conversations were transformed into searchable, structured safety intelligence with over 95% classification accuracy.
• Proactive Visibility into Leading Indicators: Plant managers gained immediate visibility into hidden safety trends, shifting the organization from lagging incident tracking to leading indicator risk management.
• 40% Increase in Targeted Coaching Effectiveness: Frontline supervisors gained clear, data-driven insights into specific coaching opportunities, driving a more proactive, measurable safety culture on the factory floor.
• Reduction in Administrative Burden: Automated synthesis and categorization cut the time safety managers spent on manual report compilation by an estimated 50%, freeing up hundreds of hours for active floor supervision.
• End-to-End Data Governance: All classification, retrieval, and query processing remain entirely inside Snowflake, keeping sensitive operational notes and enterprise data residency completely secure and compliant.
•  

Why Narwal

The success of this solution depended on accurately extracting structured meaning from highly nuanced, unstructured safety conversations. Narwal combined deep GenAI and Agent architecture expertise with advanced Snowflake Cortex AI capabilities, including Cortex Search, Cortex Analyst, and AI_CLASSIFY, to deliver a secure, scalable solution that turns everyday safety documentation into a proactive shield for the workforce.

Ready to Turn Safety Data into Proactive Workforce Intelligence?

Partner with Narwal to build a governed, AI-powered intelligence layer that helps safety teams act faster, coach smarter, and get ahead of risk before it becomes a recordable incident.