An industrial advisor that turns a site's live telemetry into actionable recommendations. It is currently in development — this page outlines what is deployed today and what lies ahead.
Every platform that collects operational data eventually faces the same fundamental question: what does it do with it? Traditional dashboards merely displace the problem rather than solve it — a human still has to monitor, interpret, and decide. The Copilot is engineered to close that operational gap: reading real-time events across an industrial site and alerting decision-makers to what warrants immediate attention and how to resolve it.
That capability cannot exist without an uncompromising data foundation. Presence, zone dwell times, and productivity tracking already run live on our positioning infrastructure. What is still evolving is the layer above: understanding the specific nature of ongoing work and generating reliable guidance. An AI advisor trained on thin or noisy data delivers confident, flawed advice — which in high-risk industrial environments is far worse than no advice at all.
That robust foundation is what we are deploying today. Wearable sensors with on-device neural processing units continuously collect a structured physical footprint of real operations across active industrial sites: high-precision coordinates, motion dynamics, zone dwell times, voice, and video. Every helmet deployed enriches that intelligence layer.
Hardware-level telemetry & presence
Wearable sensors with embedded NPUs continuously capture structured physical telemetry: high-precision coordinates (centimeter-class outdoors via RTK, decimeter-class indoors via UWB), IMU motion dynamics, audio, and video. Real-time presence, zone dwell tracking, and basic shift compliance are live in production today.
Live
Binding physical signals to site geometry
Inputs from UWB, RTK, and environmental sensors are fused with the site's digital zone model. Raw coordinates become verified operational facts: identifying who is operating near heavy machinery, which specific work front is staffed, and the exact duration of each deployment.
Live
Transforming motion into operational meaning
Deep learning models segment physical activity into structured operational states: active wrench time, material staging, transit between zones, or idle waiting. Tool time becomes verified digital data rather than subjective end-of-shift approximations.
In development
Catching deviations from the operational baseline
Operating on top of classified workflows, the system identifies baseline anomalies: sudden rhythm breakdowns, localized crowding, staging bottlenecks, and precursor patterns that indicate emerging safety risks or schedule slippages long before post-shift audits uncover them.
In development
Forecasting shift dynamics before they unfold
By synthesizing historical site patterns with real-time field telemetry, predictive models anticipate bottlenecks, hazardous proximity events between personnel and heavy equipment, and critical path delays — providing sufficient lead time to intervene proactively.
Next milestone
Actionable recommendations with transparent reasoning
The apex layer: an AI advisor that suggests high-impact interventions to project leadership (rebalancing crews, optimizing task sequencing, re-routing logistics) and explains the operational logic behind every suggestion. Planned integrations with BIM, MES, CMMS, and ERP provide the schedule and permit baselines necessary to distinguish genuine operational friction from expected progress.
Roadmap
The Copilot is not a shipped feature and is not part of what a customer receives when they deploy XENOM today. Real-time accounting of presence, time in zones and productivity is — the two should not be confused, and we describe the Copilot as a development direction.
We are not attaching a release date to it. The schedule depends on how quickly a sufficient dataset accumulates across live sites, which depends on deployments rather than on engineering effort alone.
What does work today is the foundation: high-precision positioning, objective time and zone data, voice screening, communication and video — each with its own published results. Those are the pages worth reading if you want to know what the platform does now.
A structured, continuously collected record of activity across live sites
Objective time in work, hazard, service and waiting zones
Position accurate enough to make zone events meaningful rather than approximate
On-device processing, so the data layer scales without shipping raw streams to a server
Voice, video and telemetry collected through the same wearable device