Vygard: The UK Lone Worker Safety Platform Built Around AI Dispatch

The Vygard dispatcher console — SOS alerts, mandatory three-step acknowledgement, Howie's sub-second briefing and the tamper-evident audit trail in one view. Lone-worker safety is one of the hardest workflows in any UK organisation. The Health and Safety Executive doesn't track lone-worker incidents as a single category, but the underlying data is unambiguous: falls from height, struck-by moving vehicles, assault at work and worker fatigue between them account for the majority of UK workplace fatalities, and lone workers are over-represented in every one of those categories. Most lone-worker incidents that end badly do so for one of three reasons. The alert never fired, because the worker was incapacitated before they could press SOS. The alert fired but landed somewhere nobody was looking — a generic group inbox at 02:00. Or the alert fired and was acknowledged, but nobody actioned the next step, because the dispatcher was cold and the context wasn't on the screen yet. Vygard is the UK-built lone-worker safety platform we partner with, and its design is built around closing all three of those gaps. It's an AI-augmented dispatch console with worker apps on iOS, Android, Apple Watch and Wear OS — plus rugged handsets and intrinsically-safe ATEX hardware for the tougher environments — wrapped around a regulator-ready audit trail. Closing gap one: the alert that never fires Most lone-worker apps assume the worker can hit a panic button. If they can, that's the easy case. If they can't — if they've fallen, been knocked unconscious, been incapacitated — the panic button is useless. Vygard runs background fall detection on phones and watches, tuned by sector. Forestry workers get chainsaw-cut auto-detection on Apple Watch with a 30-second auto-SOS escalation paired with a heart-rate spike threshold. Agricultural workers on quads get a 4-second tilt-threshold for rollover detection. Drivers get lane-drift fatigue scoring. Care workers get hostile-encounter pattern detection. On top of that, there's a welfare-ping engine — a proactive layer that asks workers if they're OK before they have to ask for help. The system scans worker state every 60 seconds against eight trigger reasons: vitals drift, livestock zone entry, weather alert polygons, crime overlays and more. When something looks off, a check-in is prompted. If the worker doesn't respond inside the window, escalation begins. Closing gap two: the alert that lands nowhere Vygard's escalation model is opinionated. When an SOS fires — by worker, by fall detection, by welfare-ping non-response — it doesn't go to a group inbox. It hits the duty dispatcher's console with a mandatory three-step acknowledgement: read, actioned, welfare confirmed. The incident cannot be closed until all three are ticked, and the audit trail is tamper-evident. The platform integrates outbound to Microsoft Teams, Slack, ServiceNow, BambooHR, Workday, Twilio for SMS and voice, MQTT for IoT and SIA-IP for ARC escalation where a 24/7 monitoring centre is part of the chain. So the alert reaches the people whose job it is to act on it, with the workflow to ensure they actually do. For the most remote workers — mountain rescue and SAR teams, agriculture and forestry on remote land, offshore oil & gas — Vygard supports Iridium satellite-fallback SOS, which kicks in when GSM coverage drops. This is the one feature that distinguishes Vygard from pretty much every other lone-worker platform aimed at urban field workforces. Closing gap three: the dispatcher who acts cold The third gap is the one most existing platforms have no answer to. An SOS at 02:00 lands on a dispatcher who's been on shift for six hours and is about to be cold-reading the incident. What's the worker's normal location? What's their last vitals reading? What sector pack are they on? Who's the next-of-kin contact? Who's the nearest peer-aid responder? By the time the dispatcher's read the screen, three minutes have gone. Vygard solves this with Howie — the AI watcher inside the platform, powered by Claude Opus 4.8. Howie's first job is the sub-second SOS briefing. The moment an alert fires, Howie composes the situation summary, the recommended next action and the priority call list inside the dispatcher's SOS modal — before the dispatcher's even finished reading the alert. Structured tool-use output, strict JSON schema, every action logged for the RIDDOR audit. Howie's other jobs are the welfare-ping decisions ("this worker is in a livestock zone, weather has just shifted, vitals are slightly elevated — is a check-in warranted?") and the anomaly flagging across the worker base. The model runs continuously rather than on alert. Regulator-ready by design Vygard ships aligned to UK lone-worker regulation. The provision is BS 8484-aligned. Adult social care deployments get CQC Reg 13 alignment. Construction and field-service work uses HSE-aligned workflows. RIDDOR-reportable incidents produce auto-PDF exports straight from the audit trail. Per sector there's additional alignment — DVSA for transport, SIA for security, MREW for mountain rescue, FISA 302 for forestry. The platform is UK-hosted in AWS eu-west-2, BYOD-friendly with no MDM enrolment required, full uninstall on offboarding, and the regulatory certification posture is published transparently on the platform's own site (SOC 2 Type II and ISO 27001 are targeted for late 2026; Cyber Essentials Plus and NHS DSPT are in flight). Twelve vertical sector packs The hardest part of any lone-worker deployment is the workflow per role. A delivery driver's risk profile is nothing like a forestry surgeon's, which is nothing like a domiciliary care worker's, which is nothing like an ATEX-zone petrochemical engineer's. Vygard ships with twelve sector packs covering: adult social care (domiciliary, residential, supported living); construction; logistics and driver fleet; security and manned guarding; mountain rescue and SAR; forestry and tree surgery; refuse and waste…