InnoGlas · Industrial AR platform · 2023
Give each device the job it does best.
Operators were losing task context as work moved between mobile, AR glasses and web.
My role: I owned research synthesis, cross-device information architecture, interaction design and prototyping. I partnered with Roy and Mike on development.
Key decision: Assign readiness and recovery to mobile, in-equipment guidance to the glasses, and remote assistance, recordings and procedure management to the web.
What happened: A comparison with five industrial workers informed priorities for clearer connection states and more dependable overlays.
The field task is the centre of the experience. This web view gives an expert live context when the operator needs help.
Design the handoffs around the operator’s task.
The operator needs to know that the device is ready, follow a procedure at the machine and reach an expert if a step fails. If those moments live in disconnected interfaces, the operator loses context at the point where it matters most.
Define the workflow before dividing it across devices.
Market review, operator interviews and survey inputs
I synthesised competitor audits, regional survey input and operator interviews around setup, task interruption, training and access to expertise. I mapped preparation, guided work, support and follow-up as a connected sequence. The product question was which device should own each task, and what context needed to follow the operator.
Design implication: I mapped preparation, guided work, support and follow-up as one sequence. The key product question became which device should own each decision, and what context must follow the operator between them.
Research archive · Scroll to review the evidence behind the workflow.























Make each surface’s responsibility explicit.
Journey mapping and information architecture
Mobile confirms readiness, supports task entry and provides recovery when setup fails. AR glasses present guidance where the operator is working, close to the equipment. Web coordinates remote assistance, recordings and procedure management. The aim was to reduce the burden on the wearer while keeping recovery and expert support accessible.
Decision artifacts · Scroll to review how the cross-device scope was defined.





Make connection states part of the experience.
Connection states and recovery paths
I designed disconnected, connecting and ready states so the operator could understand the system before entering a procedure. The companion keeps task entry and support visible. When setup is interrupted, the interface needs to explain the state and provide a recovery path.



Design rationale: surface readiness before task entry, and leave the support path visible if setup fails.
Carry the task into guidance and support.
On-site guidance and remote expert support
The operator selects a procedure on mobile, follows guidance at the equipment and can reach an expert from the same task path. The web workspace gives the expert session context, access to recordings and tools for maintaining procedures. The handoff preserves the work in progress rather than starting a separate support task.
Choose a procedure → inspect the step → reach an expert when needed.




Locate a colleague, review the session, maintain the procedure and follow up.




Compare the workflow in the operator’s context.
Five-person formative comparison
Five industrial workers compared tasks with and without AR guidance. I reviewed time and error records alongside observation and follow-up feedback. The findings shaped priorities for clearer connection states and more dependable overlays.



A shared visual language for the AR platform
I documented reusable components and interface states in the InnoGlas UI guideline, giving the companion experience a shared language for design, implementation, and review.
Design the handoffs as carefully as the screens.
The work established a cross-device architecture, an evaluated prototype and reusable UI guidance. The design addressed the handoffs between devices as carefully as the interactions within each one.
What I would test next: task completion time and error rate in a live pilot.
Not yet measured: operational impact, which still requires a live pilot.
