Agentic EAM
Enterprise Asset Management systems have a difficult interface problem. The same platform may support organizations whose day-to-day work has little in common. A nuclear power plant and an airport, for example, have very different processes and requirements. A general-purpose interface has to cover both, which often makes routine work harder to learn and navigate.
Our two demos explore a more flexible approach: let an AI agent assemble a task-specific workspace from trusted components, while keeping people responsible for the industrial process itself.
A task-specific EAM workspace
This gives the agent a narrow, practical role. It adapts the interface to the task without asking users to navigate every part of a large EAM system, and it does so within a predefined set of UI components.
XR: spacial dashboards & interactive 3D models
The second demo carries the dynamic dashboard into an XR workspace. We recorded it on an Oculus Quest 3 in passthrough mode, so the physical surroundings remain visible while virtual panels and 3D content occupy the space around the user.
An XR dashboard is not confined to a rectangular screen. Panels can surround the operator, allowing several related views to remain visible without compressing them into one display. The same workspace can hold an interactive 3D model of a physical asset, so the operator can explore its virtual representation alongside the information connected to it.
The demo also uses HTML-in-Canvas to place an HTML-based interface inside the 3D scene. This brings familiar web UI into the spatial workspace without a headset-specific application stack. The implementation remains based on open web technologies, with WebXR and Three.js providing the spatial layer.
Technologies used
- Symbiote UI (widgets, panels, tabs, AI-chat, etc.)
- DeepSeek LLM via OpenRouter
- IBM Maximo
- UNIAPI Adapter Tool for AI-agent
- HTML-in-Canvas
- WebXR with Three.js
Why build a custom agentic layer?
Vendor-provided AI features are only one option. Many EAM deployments still run older versions, and a platform migration may be too expensive or disruptive. Security boundaries can restrict where data is sent and which models may process it. Teams may also need control over LLM providers, operating costs, or data assembled from several systems.
A custom layer can address those constraints and leave the existing EAM platform in place. Together, these demos show the areas our team can cover: EAM integration, data adapters, constrained agent behavior, component-based interfaces, HTML-in-Canvas, and WebXR delivery.