How Presien Uses Claude to Reduce Construction Safety Events by 70%

Presien spent six years building computer vision models that run directly on heavy machinery — excavators, mining trucks, bulldozers — detecting hazards like workers in blind spots, excavators operating near site boundaries, and vehicles moving without spotters. By the time the company was ready to build an intelligence layer on top of this data, it had something rare: a continuous, real-world stream of physical safety events from active worksites across the globe.

Construction and mining sites generate vast volumes of safety data — video feeds, sensor readings, detection logs — but less than 5% is ever acted upon. Before /loop, safety managers spent hours each day manually reviewing dashboards and video clips to assess risk. Briefings were generic templates, reports were compiled by hand, and there was no scalable way to cross-reference detection events against site regulations or catch compliance issues before they escalated. The data piled up while teams were already buried in review.

Presien connected its data infrastructure to Claude through MCP servers, building /loop — an agentic platform that reasons over live worksite data around the clock. When /loop identifies a cluster of high-risk detections on a site, Claude pulls the relevant events, cross-references them against the company’s safety documentation and applicable regulations, and determines what action is needed. Safety managers can also query /loop directly: “Where are my highest-risk activities across all sites this week and what should I do about it?” Pre-built Skills allow teams to trigger complex multi-step workflows — generating branded safety reports, drafting site escalation emails, blocking calendar time for review — in seconds. Presien selected Claude after testing against real worksite data: Claude executed a full multi-step sequence without needing explicit tool routing instructions, while other models required significant human rework.

Across early deployments, critical safety events fell by more than 70% within the first three months. Safety managers no longer start their day watching video clips — they receive AI-generated risk briefings built from incidents detected overnight, ready to act on immediately. One site safety manager reported that what had previously consumed the first two hours of every day — watching videos and filling in reports — has been replaced entirely by walking the site and addressing what /loop has already identified.

Presien is expanding /loop toward fully proactive operations, where both safety and operations managers begin each shift with a role-specific briefing — risks and compliance trends for safety leads, machine utilization and productivity data for operations. Through MCP, the company is also offering OEM partners — manufacturers of excavators, mining trucks, and bulldozers — the ability to embed its intelligence directly within their own platforms, making AI-powered safety monitoring a native feature of the machinery itself.