Industries / Industrial equipment and services / Industrial services / CMMS and asset management software
Use on mobile
Mobile use puts the record in the technician's hands at the machine: scanning tags, reading instructions, recording work and getting help from far away.
5
Milestones achieved, from 1974 to 2026.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1974
Bar codes with letters
David Allais and Raymond Stevens of Interface Mechanisms, later Intermec, developed Code 39 in 1974, "the first barcode symbology to use alphabetic characters in addition to numeric digits", so part numbers could be scanned rather than typed. The US Department of Defense's LOGMARS program tested it and adopted its final plan in 1981, and from 1982 about 33,000 suppliers had to mark their shipments with Code 39, by Seagull Scientific's history.
barcodeguide.seagullscientific.combarcodeguide.seagullscientific.comen.wikipedia.org
Credit
David Allais, Raymond Stevens, Intermec; US Department of Defense
2009
Instructions over the machine
At the IEEE ISMAR conference in 2009 Steven Henderson and Steven Feiner of Columbia University reported a study of a head-mounted display that showed arrows and labels marking the parts to work on inside the turret of a LAV-25 light armored vehicle. With it, users located maintenance tasks more quickly than with electronic manuals. In 2010 the researchers said the system "is a research project and has not been deployed".
Credit
Steven Henderson, Steven Feiner, Columbia University
2023
Smart glasses withdrawn
Google stopped selling Glass Enterprise Edition 2, its smart glasses for workers, on 15 March 2023 and ended support on 15 September 2023, saying "no software updates from Google are planned".
Credit
Google
2025
Headset help withdrawn
Microsoft sold and renewed Dynamics 365 Remote Assist and Guides, its software for remote expert help and step-by-step instructions on the HoloLens 2 headset, only until 1 November 2025. Both reach end of support on 31 December 2026. Microsoft had already deprecated Remote Assist for phones on 25 March 2025, and it points customers to "spatial annotation" in Microsoft Teams Mobile for "certain Remote Assist scenarios".
LessonGoogle and Microsoft both withdrew headset products for field work within four years, and Microsoft's own replacement runs on the phone.
Credit
Microsoft
2026
Phone guidance tested in noise
Mohammed Nabeel Islam, Dhuha Basheer Abdullah and Zeena Al-Kateeb of the University of Mosul, Iraq, tested a phone app that lays step-by-step instructions over the camera view and takes spoken commands such as "Next" through Vosk, an offline speech recognizer. Ten people with no mechanical experience replaced a water pump's mechanical seal with the app and ten with a paper manual. The app group took 16.85 minutes on average against 24.11, a 30.11% cut, and made 51.4% fewer errors. Over 300 trials the spoken commands were recognized 94% of the time at 50 to 60 dB and 68% at 80 to 90 dB. The paper was published in July 2026 in the Journal Européen des Systèmes Automatisés, after peer review.
LessonRecognition of even one-word commands fell from 94% to 68% in the loudest band, so voice features need testing at the noise levels where they will be used.
Credit
Mohammed Nabeel Islam, Dhuha Basheer Abdullah, Zeena N. Al-Kateeb, University of Mosul
Watch for
Work requests by voice
MaintainX says its AI Work Request lets staff "talk it through in a guided conversation, record a voice note, or type a quick description", after which its AI fills in "the title, description, location, and priority" for review. It gives no accuracy figures.
Achieved whenan outside test reports how often the drafted fields were right, including requests recorded in noisy plant areas.
Claimed by
MaintainX, September 2026
Open
01
Voice close-outs in noise
A test of turning a technician's spoken account of a finished job into a complete work order (what failed, why, what was done, parts used, time taken) in the noise of a working plant, with speech recognition both in the cloud and on the phone.
The briefCredit · unclaimed02
Phone help, measured
A repeat of the 2009 Columbia study with today's devices and trained technicians. They do ten standard tasks on a training rig guided three ways: by paper instructions, by a remote expert on a phone video call who can draw on the screen, and by a phone app that lays instructions over the camera view. Times and errors are recorded for each.
The briefCredit · unclaimed
Open 01 · the brief
Voice close-outs in noise
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A test of turning a technician's spoken account of a finished job into a complete work order (what failed, why, what was done, parts used, time taken) in the noise of a working plant, with speech recognition both in the cloud and on the phone.
Why it is still open
Voice features now ship in CMMS products without published accuracy. The 2026 University of Mosul study measured only one-word commands, recognized 94% of the time at 50 to 60 dB and 68% at 80 to 90 dB, and searches found no public test of spoken close-outs at plant noise levels.
Built with what exists
recordings made with consent on ordinary phones; ElevenLabs Scribe v3 for cloud transcription (see the Note "Fewer word errors"); whisper.cpp, built on the ggml library, for transcription on the phone without a network (see the Note "A first stable number for local AI"); a sound level meter; a language model to fill the work order fields.
First step
Record 30 close-outs, read from real past work orders, in a quiet room and beside running equipment, and transcribe them by hand as the answer key.
Done looks like
300 recorded close-outs in three noise bands (below 70, 70 to 85, and above 85 dB(A)), published with the word error rate and the share of work order fields filled correctly for each system in each band.
Open 02 · the brief
Phone help, measured
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A repeat of the 2009 Columbia study with today's devices and trained technicians. They do ten standard tasks on a training rig guided three ways: by paper instructions, by a remote expert on a phone video call who can draw on the screen, and by a phone app that lays instructions over the camera view. Times and errors are recorded for each.
Why it is still open
Microsoft's replacement for its headset software is "spatial annotation" in Teams Mobile for "certain Remote Assist scenarios". The 2026 University of Mosul study measured a phone overlay app on one task with 20 people who had no mechanical experience. Searches found no published measurement of remote expert help by phone, or of trained technicians across several tasks.
Built with what exists
an unpowered training rig, such as a pump and gearbox for assembly practice; Microsoft Teams Mobile; Apple ARKit or Google ARCore for the overlay app; video and a timing protocol.
First step
Write the 10 task cards for one rig and time three volunteers doing them from paper.
Done looks like
At least 20 technicians doing 10 tasks under each condition in a balanced order, with task times, errors and calls for help published.