Where AR helps in manufacturing
AR helps most where a worker must follow a complex procedure, find the right part or point on a machine, or get help from someone who is not on site. The main use cases are:
- Guided work instructions: step-by-step assembly or changeover steps overlaid on the actual equipment.
- Maintenance and repair: point the device at a machine to see its history, manuals and the procedure for the fault.
- Remote expert help: a technician shares their camera view and a specialist draws on it live.
- Training: new staff practise on real equipment with guidance, or on a virtual copy when the real one cannot be stopped.
- Quality inspection: overlay the design on the finished part to check that holes, welds or components are where they should be.
- Picking and kitting: show which bin and how many.
- Layout planning: place a full-size model of a new machine or cell on the floor before it arrives.
- Sales and customer support: show buyers your equipment at scale, or guide a customer through a fix. That overlaps with AR marketing apps.
Choosing the first use case
Choose the first use case by looking for a task that is frequent, error-prone and currently depends on a paper binder or one experienced person. Score your candidates with this table.
| Question | Good sign | Poor sign |
|---|---|---|
| How often is the task done? | Daily or weekly, by many people | Once a year |
| What does a mistake cost? | Scrap, rework, downtime or a safety risk | Minor and easy to fix |
| Is the knowledge written down? | Procedures exist and are kept current | It lives in one person's head |
| Does the worker need to see the equipment while reading? | Yes, location on the machine matters | A checklist on a tablet would do |
| Do you have 3D models or good photos of the equipment? | CAD is available | Nothing exists |
| Can you measure the result? | Task time, error rate, call-outs | No baseline |
Be honest about the fourth row. Many "AR" projects are better served by a plain digital work instruction app with photos and video, which costs less and works on any device. You can add AR later to the steps that need it.
Tablet, phone or headset
Start with tablets or phones unless the task truly needs both hands at every step. They are cheap, familiar, easy to manage and good enough for most overlay work.
- Tablets and phones: lowest cost and risk. They use ARKit on iOS and ARCore on Android. The drawback is that one hand holds the device, which you can partly solve with mounts and stands at the workstation.
- Head-mounted displays and smart glasses: hands-free, which suits assembly and repair. They cost more, the product market changes quickly, battery life and comfort over a shift are real limits, and they need their own safety review.
- Fixed screens with cameras: at a single workstation, a mounted camera and monitor can give an AR view with nothing to wear or hold.
Because headset products come and go, build the content and logic so they can move between devices. Keep procedures, 3D models and data in your own backend, not locked inside one vendor's device. Our overview of AR app development explains the underlying frameworks.
What makes the plant floor hard
The plant floor is harder than a living room because of lighting, surfaces, connectivity and safety. Test for these early, on site.
- Tracking: shiny metal, repeated shapes and low light make it harder for the camera to lock onto a machine. Printed markers or QR codes fixed to equipment are a simple, reliable anchor. Tracking against the 3D model of the object is possible but takes more tuning.
- Connectivity: Wi-Fi is often patchy between machines. Cache procedures and models on the device and sync later.
- Safety: a worker looking at a screen is not looking at a forklift. Define where and when the app may be used, keep overlays out of the central field of view when moving, and involve your health and safety team from the start.
- Gloves, noise and dirt: use large controls, voice commands where noise allows, and rugged cases.
- Hazardous areas: some zones only permit specially certified devices. Check before you choose hardware.
- IT and OT security: devices on the plant network need management, updates and network separation from control systems.
Content and system integration
The lasting value comes from connecting AR to your real procedures and machine data, so plan content and integration as carefully as the app.
- CAD conversion: engineering models are too heavy for mobile devices. You need a repeatable process to simplify them and protect sensitive design detail.
- Authoring: your engineers should be able to create and update procedures without a developer. Decide whether to build a simple authoring tool or use an existing one.
- Version control: the worker must always see the approved revision, with a record of which revision was followed.
- Maintenance system: pull work orders and asset history, push completed steps, readings and photos.
- Machine data: show live values from sensors next to the machine, read-only. See IoT app development.
- Identity and roles: single sign-on, and records of who did what.
This is mostly backend work. Our backend development page explains how we approach APIs, sync and admin tools, and enterprise app development covers rollout inside larger organisations.
What drives the cost
Cost depends on scope, so treat any number given without a site visit or a detailed brief with caution. The main drivers are:
- Number of procedures and machines in the first release.
- 3D content: whether CAD exists and how much conversion and clean-up it needs.
- Tracking method: markers are cheapest, model-based tracking costs more.
- Hardware: tablets you already own, or new headsets plus device management.
- Authoring tools for your engineers.
- Integrations with maintenance, quality, ERP and machine data systems.
- Offline support and sync.
- Remote expert video, which usually relies on a paid video service.
- Security review and on-site testing.
Compare a custom build with off-the-shelf industrial AR platforms too. A platform can be quicker for standard work instructions. Custom makes sense when you need deep integration, control of your data, or a product you can also offer to your own customers.
How to run a pilot
Run a pilot on one line or cell with one use case, a clear baseline and a fixed end date.
- Pick the use case with the table above and record the current task time and error rate.
- Walk the floor with the developer. Check lighting, Wi-Fi, surfaces and safety rules.
- Prototype with markers and tablets first. Prove the tracking on the real machine.
- Write the content with the operators who do the job. They know where the binder is wrong.
- Run the pilot with a small group across different shifts and experience levels.
- Compare against the baseline, collect operator feedback, and decide: expand, change or stop.
A stop decision after a cheap pilot is a good outcome too. If you want a second opinion on a use case, send us a short description of the task and the equipment. We will tell you whether AR is the right tool, and if it is, send a fixed-scope quote for a pilot. You can also read how we run projects.