Industries / Construction and property / Janitorial and facility services
Building checks and small repairs
Building checks are the routine rounds and small jobs that keep a building working between bigger projects: looking for faults, changing lamps and filters, flushing little-used taps and logging what was done.
5
Milestones achieved, from 1990 to 2020.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1990
A standard maintenance library
SFG20, a standard maintenance specification for building engineering services, was created in 1990 by the Building Engineering Services Association (BESA). By its publisher's count it now holds thousands of maintenance schedules covering more than 110 kinds of asset, each saying which checks the asset needs and how often. The UK government's Property Data Standard of February 2023 uses SFG20 codes to classify the assets that departments maintain and to rate how critical each one is.
Credit
Building Engineering Services Association
2014
Flush the little-used taps
Britain's Health and Safety Executive (HSE) first published its code of practice for controlling Legionella bacteria in water systems, known as L8, in 1991, and its fourth edition in 2013. The technical guidance that goes with it, HSG274 Part 2 (2014), says that weekly flushing of taps and showers that are not in regular use, for several minutes, can greatly reduce the risk of Legionella growth, and that once started it "has to be sustained and logged". It also asks that hot water reach 50 °C within one minute at the outlets.
Credit
Health and Safety Executive
2017
Hospitals need water plans
ASHRAE, the US society of heating and air-conditioning engineers, published Standard 188 on 26 June 2015, calling for water management programs in large or complex buildings to reduce the risk of Legionnaires' disease. On 2 June 2017 the Centers for Medicare & Medicaid Services (CMS) told US hospitals, critical access hospitals and nursing homes that they must have water management plans that consider the standard and a CDC toolkit, assess where Legionella could grow, and record test results and the corrective actions taken; facilities that cannot show such measures risk being cited. CMS noted that reported cases rose 286% in the US from 2000 to 2014.
Credit
ASHRAE; Centers for Medicare & Medicaid Services; CDC
2019
Fault software measured
Fault detection and diagnostics (FDD) software watches a building's control data and flags faults, such as a valve left open or a fan running at night. Researchers at Lawrence Berkeley National Laboratory reported in November 2019 that 26 organizations using FDD across 550 buildings and 97 million square feet achieved median savings of 8%. Users' median costs were $8 per monitored point for the software, $2.70 per point a year in recurring fees and $8 per point a year in labor.
Credit
Guanjing Lin, Hannah Kramer and Jessica Granderson, Lawrence Berkeley National Laboratory
2020
Empty buildings, idle water
When COVID-19 closed buildings in 2020, water sat still in their pipes. In 73 hotels tested on reopening, those closed for more than three months saw the share of hot-water samples positive for Legionella rise from 6.7% in 2019 to 14.0% in 2020, by Molina and colleagues' count. In one building studied by Rhoads and colleagues, weeks of low use brought no change in Legionella pneumophila in one period, a more than 10,000-fold rise in another and a fall in a third, and precautionary flushing before people returned raised it more than tenfold. An analysis of a US lodging company's records by Kunz and colleagues found more positive samples during the pandemic and said the company's flushing rules may have limited the rise.
LessonA closed building still needs its water checks, and flushing done the wrong way can stir up the bacteria it is meant to clear.
Credit
Juan José Molina and colleagues; William Rhoads and colleagues; Jasen Kunz and colleagues
Watch for
Robot rounds in data centers
FacilityOps says its inspection robot, which carries fisheye, zoom, close-up and thermal cameras, has run rounds since 28 July 2026 at the Cossey Road Solution Center (CRSC) Data Center in Houston, powered by Solutions IS. It reports 19 autonomous runs and more than 1,600 observations in the first three weeks, and says the thermal camera spotted a failed condenser on a cooling unit 4.5 hours before the facility's own sensors alarmed, a reading a handheld check confirmed to within 1.1 °F.
Achieved whenthe data center's operator or another outside party publishes months of the robot's findings checked against technicians' rounds, with misses and false alarms.
Claimed by
FacilityOps, October 2026
AI triage of work orders
Fexa, which sells facility maintenance software, says its FexaAI Work Order Agent sorts incoming work orders and passes on only those that need an outside contractor. It says customers with more than 100 locations avoid 484 work orders a month, about 7% of their volume, and have cut triage work by 30%.
Achieved whena customer or an auditor publishes work-order and invoice counts from before and after, including requests the agent wrongly held back.
Claimed by
Fexa, June 2026
Open
01
Check the flushing logs
A study that clips small temperature loggers to the pipes just before little-used taps and showers in a consenting building, with no change to the plumbing, and compares what they record with the paper flushing log for twelve weeks: how many logged flushes show a temperature change, how long each lasted, how many flushes happened without being logged, and whether hot outlets reached 50 °C within one minute.
The briefCredit · unclaimed02
Score the help-desk AI
An open test set of 1,000 building maintenance requests, written the way occupants and staff write them, each labeled by two facility managers working apart with the trade it needs (plumbing, electrical, heating and cooling, doors and locks, cleaning) and its urgency. Then score AI triage tools on trade, urgency and how many urgent requests, such as a leak or no heat in winter, they mark as routine.
The briefCredit · unclaimed
Open 01 · the brief
Check the flushing logs
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A study that clips small temperature loggers to the pipes just before little-used taps and showers in a consenting building, with no change to the plumbing, and compares what they record with the paper flushing log for twelve weeks: how many logged flushes show a temperature change, how long each lasted, how many flushes happened without being logged, and whether hot outlets reached 50 °C within one minute.
Why it is still open
HSE guidance asks for weekly flushing of little-used outlets to be logged. Firms such as PTSG now sell remote temperature sensors for outlets and tanks, pitched in part as a way to save the water that manual flushing uses, but a search found no published comparison of paper flushing logs with measured outlet use. What flushing achieves is itself unsettled: in a pilot plumbing rig, flushed showers held less Legionella for several days, a drop the authors call "of little practical benefit", and flushing gave no apparent benefit during long periods of low use (Meegoda and colleagues, 2023).
Built with what exists
ESP32 microcontroller boards with DS18B20 waterproof temperature probes strapped to each pipe, or ready-made USB temperature loggers; Python to turn temperature traces into flush events; the HSG274 Part 2 checklist as the reference.
First step
Ask a school or office facilities team for its list of little-used outlets and last quarter's flushing sheets, and strap loggers to five outlets for two weeks.
Done looks like
30 little-used outlets in at least two buildings logged for twelve weeks, with a published table of logged flushes, measured flushes, flushes missing from the log and hot outlets that reached 50 °C within one minute, and the method and anonymized data released.
Open 02 · the brief
Score the help-desk AI
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open test set of 1,000 building maintenance requests, written the way occupants and staff write them, each labeled by two facility managers working apart with the trade it needs (plumbing, electrical, heating and cooling, doors and locks, cleaning) and its urgency. Then score AI triage tools on trade, urgency and how many urgent requests, such as a leak or no heat in winter, they mark as routine.
Why it is still open
Pampana and colleagues released twenty years of work orders from twelve US and Canadian universities, including each order's description and the priority facility managers gave it, as a data set rather than a scored test. Sellers report their own triage results, such as Fexa's 7% of work orders avoided. An open framework from September 2025 scores language models on labeling wind turbine maintenance logs, but a search found no shared test that scores triage tools on building requests.
Built with what exists
The university data on Mendeley Data as a seed; Label Studio for labeling; low-cost text models such as Claude Haiku 5.5 or Meta's Muse Ember (see the Note "The jobs agents repeat"); trial accounts of commercial help-desk tools.
First step
Download the university data set, draw 200 descriptions, and have two facility managers label trade and urgency on their own.
Done looks like
1,000 labeled requests with the two raters' agreement reported, at least three tools scored on trade accuracy, urgency accuracy and urgent requests missed, and the requests, labels and scores published.