The Problem — pilots that succeed but never scale

The AR pilot works. Five devices, one use case, a controlled environment, a motivated team running the demo. Leadership sees it. The case is made. The budget is approved.

Then the program stalls.

Gartner and PTC research points to the same finding: 80% of enterprise XR pilots fail to reach scale within 18 months. The technology is not the reason. The infrastructure is.

A pilot on five devices with a single use case is a demonstration. It tests whether the technology can perform the task. It does not test whether the organization can sustain, manage, and scale the technology across a real production environment.

Those are different questions. Answering the first one and assuming the second is answered is where enterprise AR investments disappear.

What deployment infrastructure actually consists of

Deployment infrastructure is the set of systems, processes, and capabilities that keep an AR program running after the vendor leaves.

It includes device management — enrolling, updating, monitoring, and replacing hardware at scale. It includes content authoring — building, versioning, and maintaining AR procedures when processes change. It includes facilitator training — ensuring the people closest to the workers know how to run sessions, troubleshoot devices, and capture feedback. It includes integration — connecting AR session data to MES, ERP, or quality systems so the data is usable. And it includes measurement — tracking the metrics that tell leadership whether the program is working.

None of that is glamorous. None of it shows up in a demo. All of it determines whether the program survives contact with a real production environment.

The five operational gaps that kill enterprise AR programs

No MDM system. Mobile Device Management is the infrastructure layer that keeps every device in the deployment synchronized. Without it, devices run different software versions. Content updates push to some devices and not others. A procedure changes at the engineering level and the change never reaches the floor. The program runs on outdated instructions without anyone knowing.

No content versioning. AR procedures become stale. Processes change. Parts are revised. Equipment is updated. Without a versioning system that tracks which version of a procedure is active and pushes updates to devices, the program operates on procedures that no longer match the actual process.

No facilitator training. Workers stop using systems they do not understand and cannot get help with. If the only person who knows how the AR system works is the vendor's implementation contact, the program is one personnel change away from failure.

No feedback mechanism. Errors in AR content are not always obvious. A wrong annotation, a mislabeled component, an incorrect tolerance — these persist silently until a worker makes an error and someone connects it to the instruction. Without a structured feedback channel from floor to content authoring, errors compound.

No success metrics. Leadership loses confidence in programs they cannot measure. If the only data available is device usage logs, the program is vulnerable to budget review. Defect escape rate, assembly time, and first-pass yield are the metrics that translate AR performance into operational and financial terms.

Device management at scale — MDM, content versioning, hardware lifecycle

Every AR device in an industrial environment needs to be enrolled in an MDM system before deployment. Microsoft Intune and VMware Workspace ONE are the standard platforms for enterprise device management. Both support RealWear, HoloLens 2, and Android-based AR hardware.

MDM enables centralized content distribution — push a procedure update once and it reaches every enrolled device. It enables usage monitoring — which devices are being used, how often, and on which content. It enables remote lock and wipe if a device is lost or compromised. And it enables configuration enforcement — devices stay locked to approved apps and cannot be used for off-task purposes.

Hardware lifecycle is a planning requirement, not an afterthought. Industrial AR headsets have a service life of three to five years under production conditions. Replacement cycles need to be budgeted. Spare devices need to be available for maintenance windows. Battery degradation under daily use needs to be tracked.

The facilities that treat hardware as a one-time purchase and infrastructure as someone else's problem are the ones that end up with non-functional programs two years after deployment.

Content authoring pipelines — the invisible bottleneck

Content authoring is the highest-risk gap in enterprise AR deployment. It is also the least visible during the pilot phase.

During a pilot, the vendor builds the content. The five procedures are authored, tested, and refined on a controlled timeline. It looks manageable.

At scale, with 50 procedures across four product lines, content authoring is a production process that requires internal capability. When a process changes — and in a manufacturing environment, processes change constantly — the procedure has to be updated before the revised version reaches the floor. If that update requires a vendor ticket and a two-week turnaround, the program runs on incorrect instructions during that gap.

The solution is internal authoring capability. Someone on the team — process engineer, quality specialist, training coordinator — needs to own the authoring tool and be able to make procedure updates without vendor involvement. That requires selecting an authoring tool the team can actually operate, training the author, and building the update cadence into normal process change management.

PTC Vuforia Studio and Scope AR WorkLink are the current field-grade authoring platforms for industrial AR. Both are designed for non-developers. Both produce content that runs on major AR hardware platforms.

Change management and worker adoption as deployment variables

The single highest predictor of AR program failure in production environments is worker distrust of the device.

Distrust builds fast. If the device is complicated to put on, the interaction is unclear, the AR content does not match the actual task, or the worker sees no visible benefit in the first week — they stop using it. A supervisor who enforces usage against worker resistance is managing a failed deployment, not a successful one.

Adoption requires the system to be clearly better than what it replaces, from the worker's perspective, in the first session. That means: faster to put on than a binder to locate. Clearer than a paper procedure. Faster to complete the task than without guidance.

If those conditions are not met, the change management problem cannot be solved through communication or mandate.

Pilot the system on a task where it delivers immediate, obvious value to the worker. Build adoption from that anchor before expanding to harder use cases.

Building for durability — what a production-ready AR system requires

A production-ready AR program has a device owner, a content owner, a facilitator, and a metrics owner. Those do not need to be four different people. They do need to be defined roles with defined responsibilities.

Before deployment, document: which devices are enrolled in MDM, what the content update process is, who owns authoring, how feedback gets from the floor to the authoring team, and which metrics will be reported to leadership on what cadence.

A program without those answers documented is a pilot. A program with them is infrastructure.

Learn more about NDA's AR deployment capabilities. See how mixed reality extends AR guidance into remote expert scenarios.