Construction and infrastructure fieldwork runs on paperwork nobody reads in time. A worker fills out a safety checklist at the start of a shift. A supervisor signs it. It gets filed. If something goes wrong on the job, the form gets pulled — but by then it is too late to have done anything useful.
FYLD, the London-based AI platform founded by Shelley Copsey, was built to replace that cycle with something that actually functions as a safety system. The company has closed a $41 million Series B led by Energy Impact Partners — a New York climate and infrastructure investor — with participation from Partech’s Growth Impact Fund. The round follows 82% year-over-year revenue growth in 2025 and brings FYLD’s total funding to a level that positions it for serious North American expansion.
The product is, at its core, a swap: instead of filling out a form, a field worker shoots a short video of the work site. FYLD’s AI analyses that video in real time, identifies safety hazards, quality issues, and delivery risks, and surfaces them to supervisors and project managers before they escalate. The output is not a report to be filed — it is an alert to be acted on.
“The momentum we’re seeing reflects a clear shift,” Copsey said at the Series B announcement. “Infrastructure leaders know reactive management of their frontline workforce has never worked at scale.”
The Case for Video Over Forms
The intuition behind FYLD’s approach is worth examining, because it runs against the instinct of most safety software, which tries to make forms faster or more structured rather than replace them.
Forms have a fundamental problem: they capture what workers choose to write down, at the moment they fill them out, in the language that is available to them in the field. A worker who notices that the trench next to their task looks deeper than it should may not have the vocabulary to flag it precisely on a form. They will almost certainly capture it on video — because pointing a camera at a problem is the natural way humans communicate about physical space.
FYLD’s AI models are trained to identify specific risk categories from video: fall hazards, PPE compliance, equipment proximity risks, excavation conditions, unstable formwork. The system is tuned to construction and infrastructure contexts specifically, which means it is looking for the right things rather than attempting general-purpose scene understanding.
The headline result is a reduction in serious worksite injuries and incidents of up to 48% across the customer base — a figure that, if it holds across diverse project types and geographies, represents one of the more significant demonstrated safety outcomes in construction technology to date.
Who Is Using It
FYLD’s customer roster has been built around the largest infrastructure contractors in North America and Europe. Kiewit Corporation, Quanta Services, and Emery Sapp & Sons represent the North American civil and energy infrastructure market. Ferrovial — the infrastructure group whose projects include major road and rail programmes across the UK and US — is a longstanding partner. Sulzer, the Swiss industrial services company, brings industrial construction exposure.
These are not pilot customers. These are tier-one contractors deploying FYLD on active projects with real safety liability. The fact that Kiewit and Quanta — companies whose safety cultures are among the most scrutinised in the industry — have committed to the platform signals a level of field validation that marketing claims cannot substitute for.
Matthias Dill of Energy Impact Partners was direct about the investor thesis: “FYLD is making this possible through their video-based AI tools that uplift fieldworker capacity and safety and provide direct ROI.” Rémi Said at Partech added that FYLD “is tackling large-scale infrastructure projects, a category that has been underserved by technology for decades.”
Why Infrastructure Is the Right Beachhead
Infrastructure construction — roads, energy networks, water systems, large civil works — has characteristics that make it distinctly difficult for generic construction software. Projects are geographically distributed rather than concentrated on a single job site. Workers move through varying conditions daily rather than operating in a fixed building envelope. The crews are often from multiple subcontractors, across multiple geographies, with varying training and documentation habits.
Those conditions make a form-based safety system essentially unenforceable. A supervisor cannot be everywhere at once. A field worker on a remote section of pipeline cannot meaningfully communicate a hazard through a written checklist before someone walks into it.
Video changes that calculus. A thirty-second clip shot on a phone captures spatial information that a form cannot. FYLD’s AI extracts that information and puts it in front of the person who needs to act on it, in time to act on it.
The Series B capital will fund continued scaling in North America, where FYLD added its most significant new customers in 2025, and further development of the AI models that underpin the platform’s risk detection. Copsey has framed the company’s trajectory not as incremental improvement on existing tools but as a structural shift in how infrastructure fieldwork is managed — from reactive to predictive, from paperwork to intelligence.
