Construction employs roughly 8% of the private-sector workforce and accounts for about one in five of its fatalities. That ratio has persisted for decades.
The sector is also running short of people. The Associated Builders and Contractors estimated it needs approximately 499,000 additional workers in 2026 to meet demand, with over 80% of contractors reporting difficulty hiring. The U.S. construction market reached $2.2 trillion in 2025, according to Construction Coverage, and the pressure to deliver projects faster with fewer people is pushing firms toward automation.
Artificial intelligence (AI) vendors are answering. Construction and infrastructure companies are deploying generative AI and computer vision tools directly into job-site operations to detect safety risks, automate compliance monitoring and coordinate field teams in real time.
Ferrovial Deploys More Than 30 Agents Across Workflows
Ferrovial, the global infrastructure company behind highways, airports and tunnels across North America and Europe, is among the furthest along. The company deployed more than 30 AI agents into daily workflows using DXC Technology’s AI Workbench platform, built on Microsoft Azure. The agents make real-time decisions across field operations, safety monitoring, regulatory impact assessment and competitive analysis for Ferrovial’s 24,000 employees.
DXC announced the platform in April 2025 with Ferrovial as its anchor client. The deployment represents one of the more concrete examples of agentic AI moving from pilot into production inside a major infrastructure operator.
DroneDeploy, whose reality capture platform is used on more than 3 million sites worldwide, unveiled three operational AI agents in October 2025: Safety AI to detect hazards, Progress AI to track construction sequences and Inspection AI for predictive asset maintenance. James Stripe, DroneDeploy’s chief product officer, said the agents process and reason about data.
Advertisement: Scroll to Continue
Computer Vision Changes the Compliance Model
The traditional approach to OSHA compliance on construction sites relies on manual safety walks. Safety officers conduct periodic inspections, document violations and file reports after incidents occur. That model has structural limits. Conventional site walkthroughs miss up to 70% of transient safety violations due to the scale and complexity of active job sites.
Computer vision systems are replacing that reactive model. AI cameras continuously scan sites for missing personal protective equipment; workers entering restricted zones and fall hazards at unguarded edges. When a violation is detected, supervisors receive an alert in real time rather than after an incident report is filed. Every violation is logged with a high-resolution image and timestamp, creating a digital audit trail for OSHA compliance and insurance verification.
Pose estimation models add a predictive layer. Rather than flagging a violation only after a worker reaches a dangerous position, these systems track body movement and joint angles to estimate fall risk seconds before it materializes.
The business case extends beyond safety. Construction firms that invest in safety programs save roughly $4 to $6 for every $1 spent, according to OSHA data. OSHA penalties for safety violations run from $15,625 to $156,259 per citation. And with 45% of firms reporting project delays directly caused by worker shortages, tools that let smaller teams monitor larger sites carry direct cost implications.
The construction industry has historically been slow to adopt new technology. The labor shortage and safety pressures are changing that calculation.