Buyer's guideAugmented Reality (AR)
Choosing AR remote assistance software for field service teams
AR remote assistance software lets a specialist see a field technician's camera view and place annotations that stay pinned to the equipment, not scribbles on a moving video frame. Choose a platform by testing it on your worst network, your own devices and your recording rules, and by checking how sessions attach to work orders. Below are the criteria, a scorecard and a pilot design.
On this page
- When a video call is enough and when anchored AR earns its cost
- Capabilities to test hands-on during any vendor demo
- Phones, tablets and smart glasses for remote sessions
- Planning for weak signal, dead zones and no signal at all
- Setting the recording and consent policy before the first session
- Connecting sessions to work orders, asset history and the knowledge base
- Security and procurement questions for every shortlisted vendor
- A pilot that produces a decision rather than a demo
- Scorecard for comparing remote assistance platforms
- Questions and answers
- Sources
When a video call is enough and when anchored AR earns its cost
Most remote help starts as an ordinary video call: the technician points a phone at the problem and an expert talks them through it. That works when the fault is easy to see and the technician can hold the phone and work at once. It breaks down when the expert must point at one of a dozen identical connectors, the camera keeps moving or the job needs both hands.
Dedicated remote assistance software adds three things. Annotations are attached to the scene through the device's tracking, so an arrow on a valve stays on that valve as the camera moves. The expert can freeze a sharp frame, mark it up and send it back. And the session belongs to a job, with attendees, evidence and outcome stored against the work order. If none of that matters, a standard meeting tool with good mobile support may be the right purchase.
Typical fits are first-time fixes on unfamiliar equipment, second opinions before sign-off and supervised onboarding. ColdAI's augmented reality practice treats remote expert annotation as one overlay pattern among several1.
Capabilities to test hands-on during any vendor demo
Ask to run each test yourself, on your own device, rather than watching a presenter. Record what fails as carefully as what works.
Phones, tablets and smart glasses for remote sessions
| Factor | Phone | Tablet | Assisted-reality glasses | See-through AR headset |
|---|---|---|---|---|
| Hands-free work | No; one hand holds the device | No; usually propped or mounted | Yes | Yes |
| Annotation anchored in 3D | Yes, in the screen view | Yes, on a larger screen | Often limited to marked-up stills on a small display | Yes, in the wearer's field of view |
| Bright light and outdoor work | Screen glare is common | Screen glare is common | Varies by display; test on site | Displays can wash out in sunlight; test on site |
| Fit with hard hats and safety eyewear | Not affected | Not affected | Many models offer helmet or eyewear mounts | Check fit and any eye-protection rating the site requires |
| Rollout and support effort | Lowest; uses devices already issued | Low | Moderate; a new device class to manage | Highest; fleet management, fitting and training |
Device behavior varies by model and changes often. Treat the table as a list of questions to test, not as a verdict on any product.
Planning for weak signal, dead zones and no signal at all
Field sites are where networks are worst: plant rooms, basements, rooftops, rural substations and vessels, none of which an office demo reproduces. Ask how the platform adapts as bandwidth falls: does it cut frame rate before resolution, switch to audio plus stills and recover without dropping the call?
Then plan for no connection. Live assistance cannot run offline, but the work around it can: procedures, drawings and asset history downloaded before the visit, photos and notes queued for upload. Agree a fallback, such as a callback once the technician reaches signal, or a short recorded clip answered with marked-up stills, and test it in the pilot.
On customer premises, check whether policy allows a camera stream to leave the site. ColdAI designs AR for outdoor light, gloves, noise and patchy connectivity, so these questions come first in our projects1.
Setting the recording and consent policy before the first session
Remote sessions stream video of people and places that did not sign up for it. Under the General Data Protection Regulation (EU) 2016/679, recordings of identifiable people are personal data, needing a lawful basis, clear information and limited retention2. This is not legal advice; agree the policy with your data protection officer.
- If
Sessions take place on customer premises.
ThenTell the customer in advance that video will be streamed, and offer an audio-and-stills alternative if they object.
The customer controls the site and may have its own rules on cameras and restricted areas.
- If
Other workers or members of the public may appear on camera.
ThenBrief technicians to frame equipment, not people, and prefer platforms that can mask regions in recordings.
Bystanders were not informed beforehand, so capture as little of them as possible.
- If
Recordings will be reused for training.
ThenTreat training as a separate purpose: inform staff, edit clips before publishing and give the library its own retention rule.
A clip recorded to fix one fault is not automatically fit for every new starter.
- If
Work happens where an explosive atmosphere may occur.
ThenUse only devices certified for that zone, or keep the device outside the area.
In the EU, equipment for such areas falls under Directive 2014/34/EU, and that certification applies to the hardware, not to the software running on it3.
Connecting sessions to work orders, asset history and the knowledge base
Integration decides whether remote assistance becomes part of the job or stays a side channel. The minimum is a session started from the work order in the field service management system, with the asset, site and fault already attached.
When the session ends, annotated stills, attendees and a short outcome note should flow back to the work order and the asset's history in the maintenance system, where the next technician will find them. Ask whether this uses standard connectors or custom API work, and who maintains it through upgrades.
Recurring fixes captured in sessions can later become AR work instructions, aligned with a method from our AR tracking comparison, once someone has reviewed them.
Security and procurement questions for every shortlisted vendor
A pilot that produces a decision rather than a demo
Pick the job types
Choose two or three job types with frequent escalations and a known cost of repeat visits. Easy jobs flatter any tool.
Choose contrasting sites
Include a location with poor coverage and a customer site with strict rules, so hard cases are tested early.
Record a baseline
Capture repeat visits, time to resolution and expert time for the chosen jobs, using measures operations already trusts.
Train both sides
Experts need practice too: annotating clearly, freezing the frame and closing with a usable note.
Keep going until the tool is routine
A week of enthusiasm proves little. Run until sessions feel ordinary, with a comparison group working without the tool.
Decide against thresholds agreed upfront
Agree beforehand what result justifies rollout, a second pilot or stopping.
Scorecard for comparing remote assistance platforms
| Criterion | Strong answer | Weak answer |
|---|---|---|
| Annotation anchoring | Marks stay on the part as the camera moves | Marks drawn on the video frame drift off target |
| Weak-network behavior | Degrades to stills and audio, recovers on its own | Call freezes or drops without warning |
| Work order integration | Session opens from the job and writes evidence back | Links and files copied across by hand |
| Recording controls | Central defaults, masking and per-team retention | Each user decides what to record and keep |
| Security evidence | Current reports and a clear data-flow map | Marketing statements only |
Questions and answers
Do technicians need smart glasses for AR remote assistance?
No. Phones and tablets handle most remote sessions, and they are already issued, supported and familiar. Glasses help when the technician needs both hands throughout the job, works at height or uses remote help daily. Start on phones, measure who uses the tool and how often, then give glasses to the people for whom hands-free work changes the job.
Can remote assistance sessions be recorded and reused for training?
Most platforms can record, but training is a different purpose from fixing the fault in front of you. Inform the people recorded, review clips before reuse, remove footage of bystanders or sensitive areas and give training clips their own retention rule. Short, edited clips of a recurring fix usually teach more than full recordings.
What happens to remote assistance when the technician has no signal?
Live assistance needs a connection, so plan around the gap. Download procedures, drawings and asset history before the visit, let the technician record clips and notes offline for later upload, and agree a fallback such as a callback once signal returns. Test that fallback during the pilot, not after rollout.
Should equipment manufacturers be allowed to join our remote sessions?
Often yes, for warranty work or rare faults, but treat it as guest access. Use expiring links, limit what the guest can see and record, check that your customer agreement allows a third party to view the site, and log every guest session against the work order.
Sources
- Augmented Reality (AR): ColdAI's approach to object-aligned AR — ColdAI
- Regulation (EU) 2016/679 (General Data Protection Regulation) — EUR-Lex · checked 10 October 2026
- Directive 2014/34/EU on equipment and protective systems intended for use in potentially explosive atmospheres — EUR-Lex · checked 10 October 2026