The most compelling AI safety story in any UK industry

In the debate about AI in UK industry, few data points are as striking as this one: AI-powered collision avoidance systems fitted on more than 5,000 items of plant machinery in the UK have recorded zero reported accidents. (Beale & Co)

In construction — historically one of the highest-risk industries in the UK — zero accidents on 5,000 machines is an extraordinary outcome. It is the kind of result that should be making front pages. Instead, it sits largely unreported outside specialist construction press, while more abstract debates about AI risk and regulation dominate the mainstream conversation.

The construction sector's AI story is rarely told as a safety story. It should be.

The scale of the safety problem AI is solving

Construction accounts for a disproportionate share of UK workplace fatalities. HSE data consistently shows construction representing approximately 4% of the UK workforce but accounting for around 28% of fatal injuries to workers. Plant machinery — excavators, telehandlers, dumpers, and other mobile equipment — is a major category in serious and fatal construction incidents, particularly involving struck-by accidents between plant and pedestrians on congested sites.

AI collision avoidance addresses this specific risk. Camera and sensor systems fitted to plant machinery use machine learning to identify pedestrians, other vehicles, and obstacles in real time, providing audible and visual warnings to operators and, in some systems, automatic slowdown or stop commands when a collision is imminent.

The technology is not new — proximity warning systems have been available for years — but AI-powered systems represent a significant capability upgrade: they distinguish between humans and objects, reduce false alarms that cause “alarm fatigue,” and adapt to different site environments and lighting conditions in ways that rule-based systems cannot.

The 40–50% incident reduction finding

Beyond the zero-accidents headline from collision avoidance systems specifically, broader AI-powered site monitoring is delivering significant incident reductions.

Companies implementing AI-powered continuous site monitoring report 40–50% reductions in incidents. (WCCTV) These systems use AI-analysed CCTV to automatically detect safety non-compliance — workers without appropriate PPE, exclusion zone breaches, unsafe behaviours — in real time, without requiring a human to watch feeds.

The 24/7 nature of AI monitoring is particularly important on large sites with multiple contractors and shift patterns. Human-monitored safety typically focuses on the situations that are observed and reported. AI monitoring is consistent, objective, and continuous — it catches what tired or distracted human observers miss.

The “Big Brother” challenge and what the HSE said

The expansion of AI monitoring on construction sites has not been without resistance. Workers and trade unions have raised legitimate concerns about the use of surveillance technology to monitor workers — concerns about privacy, trust, and the use of monitoring data for disciplinary rather than safety purposes.

HSE CEO Sarah Albon addressed this directly at an AGM, stating that monitoring workers was “not 'Big Brother' providing it was used for purposes of safety and that consultation before use had taken place.” (British Safety Council)

This formulation establishes the principles the HSE is applying: purpose limitation (safety, not performance management or surveillance) and prior consultation (workers must be informed and engaged before monitoring begins). Firms implementing AI monitoring should treat these not as optional best practice but as the regulator's articulated expectations.

The consultation requirement is particularly important. Top-down implementation of AI monitoring without worker buy-in does not just create morale problems — it actively undermines the safety outcomes the technology is meant to deliver. Workers who distrust or resent monitoring systems find ways to work around them, or fail to report near-misses for fear of consequences.

The manager-worker perception gap: why it matters

Rapid Global's 2025 UK construction survey found a stark gap in how managers and frontline workers perceive AI safety readiness:

51% of managers believe their business is ready for AI safety tools

20% of frontline workers feel the same

(International Fire & Safety Journal)

This 31-percentage-point gap is not a minor disagreement about implementation readiness. It signals a disconnect that, if unaddressed, will undermine AI safety implementations. Workers who do not believe in or understand the technology they are working alongside will not integrate it effectively into their behaviours.

Morgan Sindall's experience provides a useful model for what effective implementation looks like. Director Dwayne Florant described the lesson clearly: “If you involve the site teams early — in testing and choosing the tools — they take ownership. That's when you start to see real behavioural change.” (Building)

This is not merely good people management — it is the operational requirement for AI safety tools to deliver their potential. A collision avoidance system that operators learn to work around because they distrust it is no safer than no system at all. Worker co-design of AI safety implementation is a safety intervention, not just an HR nicety.

Specific AI safety applications gaining traction in UK construction

Collision avoidance and proximity detection: Fitted to excavators, dumpers, telehandlers, and other plant. Camera and radar-based systems identify pedestrians and vehicles in danger zones. Fitted on 5,000+ UK machines with zero reported accidents. (Beale & Co)

PPE compliance monitoring: AI-analysed CCTV that identifies workers without hardhats, hi-vis vests, or other required PPE. Alerts can be generated in real time to site managers via mobile notifications, enabling immediate intervention.

Exclusion zone monitoring: AI systems that define virtual boundaries around dangerous areas (excavations, crane operating areas, live electrical equipment) and alert when these are breached.

Driver behaviour monitoring: Systems that track plant operator behaviour — speed, harsh braking, phone use — and flag patterns associated with fatigue or distraction.

Predictive safety analytics: AI systems that analyse historical incident data, near-miss reports, and site conditions to predict elevated risk periods and locations, enabling proactive rather than reactive safety management.

The Building Safety Act compliance dimension

The Building Safety Act's Golden Thread requirement — comprehensive digital records for higher-risk buildings — intersects with AI safety in an important way. Safety incidents, near-misses, and corrective actions are part of the information that the Golden Thread must capture for higher-risk buildings in construction.

AI systems that automatically log safety events, generate structured reports, and timestamp interventions create the kind of auditable, searchable safety record that the Golden Thread requires. Firms investing in AI safety monitoring for operational reasons are simultaneously building the compliance infrastructure the Building Safety Act demands. (Building Passport)

The RICS standard and what it means for safety AI

The RICS mandatory AI standard, effective 9 March 2026, requires chartered surveyors to apply professional judgment as paramount when using AI, and mandates client disclosure and explainability of AI outputs. (Beale & Co)

For AI safety systems on construction sites, this has specific implications for chartered building surveyors and project managers: they cannot simply deploy an AI safety system and treat its outputs as authoritative. They must understand the system's capabilities and limitations, maintain professional oversight of safety decisions, and be able to explain to clients and regulators how AI contributes to site safety management.

This is a higher bar than simply purchasing and switching on a system. It requires professional engagement with how the technology works and where its limits are.

The business case: liability reduction and insurance

Beyond the human cost of construction accidents, AI safety tools present a compelling commercial case. Serious workplace accidents in construction generate:

A 40–50% incident reduction — the figure reported by AI monitoring adopters (WCCTV) — translates directly into reduced exposure across all these categories. For a firm with a strong safety record seeking to maintain low insurance premiums and protect its HSE enforcement history, AI safety investment has a measurable return.

Insurers are beginning to recognise this. Firms with documented AI safety systems — and particularly those with incident data demonstrating the effectiveness of these systems — are in a stronger position in liability insurance negotiations than those relying on traditional safety management alone.

Key statistics at a glance

MarGen helps construction and professional services firms build the content authority that earns trust from both search engines and AI systems. Find out more.