WorkVis

What it does

WorkVis sells software that applies computer vision to worksite camera feeds to spot hazards and alert people while work is under way ^[1][2]. The company lists solutions for PPE compliance, vehicle collision, geofencing, entry and exit, lockout/tagout, electrical hazards, rail yards, mustering, housekeeping, confined space entry, crane safety, site inspection, hot work, fall detection and machine guarding ^[1]. WorkVis says its system first gives the worker an audio or visual alert with the corrective action, then notifies a supervisor if the worker does not respond, with escalation rules that customers can configure ^[1][8]. Alerts, recorded clips and analytics are available in a mobile app for iOS and Android ^[13].

Who it is for

WorkVis targets heavy industry: metals and steel, oil and gas, chemicals, utilities, and warehousing and logistics ^[7]. Its contact page names WorkVis.io Inc. in Pittsburgh as global headquarters and WorkVis AB in Stockholm as its European office, and both entities appear in public company registries ^[3][4][5]. The company lists Raghu Arunachalam (CEO and co-founder), Eric Hayman (CTO and co-founder) and André Dübbel (chief data scientist) as its leadership ^[2]. Public sources disagree on the founding year, so we have left it out.

How it is deployed

WorkVis says it works with existing or new surveillance camera feeds ^[2]. In its US Steel case study, cameras were installed around electrical maintenance areas and the AI models were trained on site footage ^[10]. The public material does not say where video is processed (on site, edge device or cloud), what hardware is needed, or which EHS or video management systems it integrates with. WorkVis says it holds a SOC 2 report under the AICPA Trust Services Criteria, available to customers on request, and offers face blurring and text or license plate blurring; it says it does not use facial recognition ^[9]. No pricing is published.

Notable public customers and coverage

US Steel is the only named customer case. WorkVis says a field trial began at Edgar Thomson Works in Braddock, Pennsylvania, in mid 2022, with voice alerts when workers began electrical work without the right PPE ^[10]. It reports a 95% reduction in PPE noncompliance alerts from January to September 2024 and zero electrical PPE noncompliance incidents at that site ^[10]. These are vendor-reported figures, not independently verified, and the case study itself notes that part of the drop came from other factors such as a PPE change and model improvements ^[10]. Christopher Petrouski of US Steel corporate safety was scheduled to co-present the deployment with WorkVis at the 2024 NSC Safety Congress & Expo, according to WorkVis ^[11]. The Pittsburgh Post-Gazette covered the US Steel work in March 2024 ^[12]. In March 2024 WorkVis announced a growth equity investment from Nicely Road Capital, with founder Justin McElhattan, a former Industrial Scientific CEO, joining its board; no amount was disclosed ^[6][2].

Things to ask in an evaluation

• Where is video processed, what hardware sits on site, and what bandwidth and retention does each camera need?
• How was the 95% figure measured, given that it counts system alerts, and what share came from the PPE change and model updates?
• What detection accuracy and false alert rates should you expect for your PPE types and lighting, and how were they validated?
• Can you see the SOC 2 report, and is it Type I or Type II?
• Which EHS, work order or video management systems can receive alerts?
• How many deployments run outside US Steel, and can you speak to one of them?

Our take

Strengths

• A named heavy industry customer case with a described method and a stated caveat on its own figures ^[10].
• A dedicated security, privacy and AI ethics page with a SOC 2 claim and documented face and license plate blurring (vendor claims) ^[9].
• Leadership, legal entities and investor are disclosed and match public registry records ^{[2][3][4][5][6]}.

Watch-outs

• No published processing architecture, integrations, pricing or accuracy method ^[1][9].
• Public evidence rests on one customer, and the headline outcome is vendor-reported ^[10].
• No analyst, academic or user review evidence found; the mobile app has too few ratings to display ^[13].

Best fit: A US heavy industry site, such as steel or electrical maintenance areas, that wants real-time PPE and task alerts to workers and is prepared to validate accuracy and architecture in a pilot ^[7][10].