In September 2026 we surveyed 518 maintenance and operations professionals about how AI
fits into their work. Respondents ranged from technicians to executives. Three findings
shape what follows.
67.8%use AI at least occasionally
55.4%work somewhere with no AI deployed
27.8%use AI on their own at a company with none in place
What we expect of AI in 2027
Executive summary
AI adoption has already started, from the floor up.
More than a quarter of respondents (27.8%) use AI at companies that have not rolled it
out in maintenance or operations. One in ten has already built a working app, and 40.4%
of those builders are supervisors or managers. Directors and VPs are the most worried
about falling behind, with 56.0% saying that concern grew over the past year.
They are also the least likely to work somewhere with AI in regular use, at 13.7%. The
people with the authority to fund AI are the furthest from the AI their teams already
use.
Outside pressure will not set the timeline, so the operation has to.
On all five pressure questions in the survey, the most common answer was "no change"
over the past year. Wanting to work more efficiently is the biggest driver of adoption
at 42.9%.
Every outside pressure combined, from competitors to regulators, adds up to 15.3%. A
director who wants budget can build the case today from the operation's own work order
history.
The first project belongs to the people already doing the work by hand.
Asked where AI would help most, 43.1% pick paperwork, like writing up work orders and
finding past records. Only 13.7% pick diagnosing equipment, and that split holds in
every role and industry in the research. A bad work order draft takes one edit to fix.
A bad diagnosis can cost a wasted trip across the plant and the crew's trust in the
tool. Among people who built their own app, 58.7% would otherwise have kept doing the
work by hand or never done it at all, which makes them the people who know what the
first tool should do.
Chapter 1
Maintenance teams are already adopting AI on their own
Two thirds of maintenance and operations professionals (67.8%) use AI at least
occasionally, and 90.7% have used an AI tool at some point. Their employers are well
behind them.
More than half (55.4%) work somewhere that has not rolled out any AI in maintenance or
operations. Put the two numbers together and 27.8% use AI on their own at a company with
no AI in place. Every one of them started without being asked.
Figure 1 Personal use against organizational deployment
Personal AI use n=518
Most days Occasionally Tried a few times Never
All respondents
37.1 30.7 23 9.3
Organizational AI use in maintenance or operations n=518
Not using Looking into it Trying in one area Using regularly
All respondents
27.4 28 17.2 27.4
Personal use against organizational deployment
Measure
Answer
Share
Personal AI use
Most days
37.1%
Personal AI use
Occasionally
30.7%
Personal AI use
Tried a few times
23.0%
Personal AI use
Never
9.3%
Organizational AI use
Not using
27.4%
Organizational AI use
Looking into it
28.0%
Organizational AI use
Trying in one area
17.2%
Organizational AI use
Using regularly
27.4%
67.8% use AI most days or occasionally, while 55.4% are not using AI or are only looking into it. Q4 and Q5, n=518.
I can build my own apps. I can talk to my documents and it will share the information and pictures the reply requires.
When asked how they feel about AI in their own line of work, the largest group (43.8%)
said "interested, but cautious." Another 35.1% said "I am excited about what it could
do." Only 5.6% picked "I will believe it when I see it," and 6.6% said they were
concerned about what AI means for their team.
How much someone uses AI personally tracks closely with how they feel about it at work.
Among people who use AI most days, 67.2% said they were excited. That share drops to
26.4% among occasional users and keeps falling, down to 4.2% among people who have never
used an AI tool.
Figure 2 Excitement rises with personal use
67.2%
Most days
26.4%
Occasionally
7.6%
Tried a few times
4.2%
Never used one
Excitement rises with personal use
Personal AI use
Excited
Most days
67.2%
Occasionally
26.4%
Tried a few times
7.6%
Never used one
4.2%
Share who chose 'I am excited about what it could do', by personal AI use. Correlation at one point in time. Bases: 192, 159, 119 and 48 respondents.
Directors and VPs of Operations worry the most about falling behind on AI and are the
least likely to have AI in place. Among them, 49.0% are excited about AI, and 56.0% said
their concern about falling behind grew over the past year, compared with 29.7% of all
respondents. Only 13.7% work at a company that uses AI regularly in day-to-day work, the
lowest of any role in the research.
Figure 3 Directors and VPs, most worried and least deployed
Directors and VPs All respondents
Concern about falling behind increased
56%
29.7%
Company uses AI regularly
13.7%
27.4%
Directors and VPs, most worried and least deployed
Measure
Directors and VPs
All respondents
Concern about falling behind increased
56.0%
29.7%
Company uses AI regularly
13.7%
27.4%
Q7e × Q2 and Q5 × Q2. Director and VP base is 51, which the report flags as directional.
More people use AI on their own than work at companies that have rolled it out, because
trying AI alone takes far less effort. A company rollout needs a security review and a
purchase order first. For leaders, the fastest way to speed up AI adoption is to start
with the employees already using AI. They know which tasks AI handles well in the
operation, and their input can set the order of the roadmap.
Chapter 2
Set the AI timeline inside the operation
Most AI projects in maintenance are thought to stall for one of two reasons. Either
leadership has not made AI a priority, or the budget has not been set aside for it yet.
Respondents were asked whether five things had changed at their organization over the
past year.
On every one, the most common answer was "no change," chosen by between 55.1% and 72.9%
of respondents.
Figure 4 Five pressures, one answer
Decreased No change Increased somewhat Increased a lot
Competitor awareness n=515
55.1 29.5 10.7
Leadership pressure for an AI strategy n=516
9.5 57 24.4 9.1
Concern falling behind will cost business n=512
63.9 23.2
Budget for AI or automation n=513
67.6 20.5
Customer or auditor expectations n=513
72.9 14.8
Five pressures, one answer
Pressure
Decreased
No change
Increased somewhat
Increased a lot
Competitor awareness (n=515)
4.6%
55.1%
29.5%
10.7%
Leadership pressure for an AI strategy (n=516)
9.5%
57.0%
24.4%
9.1%
Concern falling behind will cost business (n=512)
6.4%
63.9%
23.2%
6.4%
Budget for AI or automation (n=513)
4.8%
67.6%
20.5%
7.0%
Customer or auditor expectations (n=513)
7.0%
72.9%
14.8%
5.3%
"No change" is the most common answer on all five rows. On teams of 26 to 100, 24.3% report a sharp increase in competitor awareness, against 10.7% overall. Q7a to Q7e, blank answers excluded.
The strongest way to get budget for AI in
maintenance is to show how much time it will save. When asked for the single biggest driver of AI adoption, 42.9% chose working more
efficiently, which outweighs every outside pressure combined at 15.3%.
Figure 5 What actually drives AI adoption
We want to work more efficiently
42.9%
Nothing is driving it right now
19.5%
Personally curious
8.5%
Competitors
6.6%
Leadership told us to
6.2%
Losing knowledge as people retire
5.4%
Other
3.5%
Tools cannot do what we need
3.3%
Customers and auditors
1.9%
Cost cutting
1.7%
Regulations
0.6%
What actually drives AI adoption
Biggest driver
Share
We want to work more efficiently
42.9%
Nothing is driving it right now
19.5%
Personally curious
8.5%
Competitors
6.6%
Leadership told us to
6.2%
Losing knowledge as people retire
5.4%
Other
3.5%
Tools cannot do what we need
3.3%
Customers and auditors
1.9%
Cost cutting
1.7%
Regulations
0.6%
Red marks the four outside pressures, which come to 15.3% together. Efficiency outdraws all of them nearly three to one. Q8, n=518.
Outside pressure makes a weaker argument. Awareness of competitors is rising at more
organizations than any other pressure, at 40.2%, yet only 6.6% name competitors as the
main reason they are adopting AI. Customer and auditor expectations are rising at just
20.1% of organizations, so a compliance argument has even less behind it. The second most
common answer to the question about drivers, at 19.5%, was "nothing is driving it right
now."
To show how much time AI will save, an operation needs two numbers it already has: the
hours the crew spends on write-ups, and the crew's loaded hourly rate. Write-up hours are
in the work order history, and the hourly rate is in the labor budget.
The cost of waiting
Three numbers an operation already has. Put yours in and the total updates as you type.
hours
Per technician. This is in the work order history.
$per hour
Wages plus burden. This is in the labor budget.
weeks
Count from the first time someone raised it.
One technician. Multiply by the crew to size it for the whole operation. Nothing you
type here is sent anywhere.
Case study
When software falls short, crews pick up the work by hand
8robotic production lines
125,000 sq fttoday, planned to exceed 500,000
Up to 150 ftline length
5 to 6 monthsin operation at the time of the interview
Axon Enterprise runs eight production lines at its facility, with robotic equipment up to
150 feet long and built from lots of small components. When a technician needed one of
those parts, the search started at a drawer.
The techs were coming in and rifling through the drawers looking for a part with like a
picture pulled up on their phone. And saying, oh, wait, is this it? Is this it? Well,
this kind of looks like it. Ah, that might be too short. I don't know. We're going to
have to go test it. And then they grab a handful of different stuff and they go see what
they have. And sometimes it comes back. Sometimes it makes it on the work order. Who
knows?
Every time a technician grabbed the wrong part and did not log it, the parts count in the
system drifted further from what was actually in the drawer.
Stewart spent about a year and a half building a fix. He called it Dewey. A technician
puts a part under a camera, and Dewey identifies it and says exactly where it is stored,
down to the drawer and slot. Every night it also sends Stewart a report of the parts used
while he was away. The idea came from watching the search repeat.
Circle K knows everything. There's no barcodes or anything involved. So why can't I do that with parts?
Stewart is far from alone. App builders in the survey were asked what they would have done
if they could not have built the app. Half (50.0%) would have kept doing the work by hand,
and another 8.7% say it never would have been built. Together, 58.7% of these apps replaced
work that was being done by hand or would never have happened at all. Only 17.4% would have
gone through a feature request or a ticket.
Figure 7 What builders would have done without building
Kept doing it manually
50%
Bought another piece of software
13%
Submitted a feature request and waited
13%
Hired an outside developer
10.9%
It never would have gotten built
8.7%
Filed a ticket
4.3%
What builders would have done without building
Instead of building
Share
Kept doing it manually
50.0%
Bought another piece of software
13.0%
Submitted a feature request and waited
13.0%
Hired an outside developer
10.9%
It never would have gotten built
8.7%
Filed a ticket
4.3%
Red is the 58.7% that replaced hand work or work that would never have happened. Navy is the 17.4% that would have gone through a feature request or a ticket. Q15, n=46.
Chapter 3
An AI rollout plan sized to what the organization can deliver
The first AI project in a maintenance operation should take paperwork off a technician's
plate.
The rollout has three levels, and each one depends on the one before it. Level 1 starts
with write-ups, the
highest-volume paperwork in the operation and the place where a
mistake takes one edit to fix. The clean records from Level 1 give Level 2 something
reliable to automate. The workflows built in Level 2 then produce the records a Level 3
system learns from.
Level 1
Write-ups that need no retraining
In every role and industry surveyed, people want AI to help with paperwork more than with diagnosis.
Level 2
Workflows with a person at each decision point
A workflow that still needs human judgment can be automated one step at a time, as long as people keep the judgment calls.
Level 3
A system that learns from its own workflows
Once workflows run on their own, their records show where work still gets stuck, which points to what to automate next.
When asked which part of their work would benefit most from AI, three of the answer
choices describe paperwork: writing up work orders and documentation, finding information
in manuals and past records, and reporting to leadership. Together they draw 43.1%.
Diagnosing equipment draws 13.7%.
Figure 8 Where AI would help most
Writing up work orders and documentation
23.2%
Planning and scheduling
17%
Finding information in manuals or records
14.9%
Diagnosing what is wrong with equipment
13.7%
None of the above
11%
Managing inventory and parts
7.7%
Compliance and audit preparation
5.4%
Reporting to leadership
5%
Training new technicians
2.1%
Where AI would help most
Part of the work
Share
Writing up work orders and documentation
23.2%
Planning and scheduling
17.0%
Finding information in manuals or records
14.9%
Diagnosing what is wrong with equipment
13.7%
None of the above
11.0%
Managing inventory and parts
7.7%
Compliance and audit preparation
5.4%
Reporting to leadership
5.0%
Training new technicians
2.1%
Blue marks the three paperwork answers, which come to 43.1%. Red is diagnosis. Q9, n=518.
In every role in this research, more people chose paperwork than diagnosing equipment as
the part of their job where AI would help most, by a wide margin. Among technicians and
maintenance staff, 43.8% picked paperwork and 19.8% picked diagnosis. Maintenance
supervisors and managers split 44.1% to 15.3%. For Directors and VPs of Operations, the
gap grows to 47.1% against 11.8%.
Figure 8.1 Paperwork against diagnosis, in every cut
Paperwork Diagnosing equipment
Directors and VPs of Operations
47.1%
11.8%
EHS and compliance
44.4%
11.1%
Maintenance supervisors and managers
44.1%
15.3%
Technicians and maintenance staff
43.8%
19.8%
Reliability and engineering
41.7%
25%
Executives and owners
26.7%
13.3%
Paperwork Diagnosing equipment
Fleet and transportation
52.4%
9.5%
Healthcare
52.3%
9.1%
Facilities and property management
50.5%
11%
Manufacturing and industrial
42%
19.3%
Food and beverage
41.8%
14.9%
Education or government
39.5%
9.3%
Energy and utilities
38.9%
22.2%
Paperwork against diagnosis, in every cut
Group
Paperwork
Diagnosing equipment
Directors and VPs of Operations
47.1%
11.8%
EHS and compliance
44.4%
11.1%
Maintenance supervisors and managers
44.1%
15.3%
Technicians and maintenance staff
43.8%
19.8%
Reliability and engineering
41.7%
25.0%
Executives and owners
26.7%
13.3%
Fleet and transportation
52.4%
9.5%
Healthcare
52.3%
9.1%
Facilities and property management
50.5%
11.0%
Manufacturing and industrial
42.0%
19.3%
Food and beverage
41.8%
14.9%
Education or government
39.5%
9.3%
Energy and utilities
38.9%
22.2%
Paperwork leads in every role and every industry. Q9 × Q2 and Q9 × Q1. Bases run from 15 to 170, so the smaller cells are directional. IT (n=10) is in the PDF appendix.
I'd want an app that reads my emails and automatically creates tasks, schedules events, and formulates documents based on requests.
The first AI project should also fit the way the crew already works. Software rarely fits
an operation as it runs today. In total, 69.3% of people bend their process to fit their
software at least sometimes, and only 11.0% say the software adapts to them. If a project
changes the crew's routine, the tool has to run reliably and the
crew has to learn a new way of doing the job. Starting with a task the crew already does the same way every shift
avoids the retraining problem, because AI only drafts what the technician was already
going to write.
Figure 9 How often teams bend their process to fit the software
Almost always Often Sometimes Rarely Almost never, the software adapts to us
All respondents n=518
15.8 18.1 35.3 19.7 11
How often teams bend their process to fit the software
Respondents
Almost always
Often
Sometimes
Rarely
Almost never, the software adapts to us
All respondents (n=518)
15.8%
18.1%
35.3%
19.7%
11.0%
69.3% bend their process at least sometimes. Q10, n=518.
What people build first is usually simple. Among builders in this research, 23.9% made
something that creates a report or summary, and 21.7% made something that watches a
condition and flags a problem. Both replace work someone was doing by hand. A good Level 1
project follows the same pattern, picking one report or check the crew already knows well,
so the time saved is easy to measure.
Figure 10 What survey respondents built
Generates a report or summary
23.9%
Monitors conditions and flags issues
21.7%
Collects data from the field
13%
Calculates something specific to our equipment
13%
Handles a scheduling or assignment rule
8.7%
Connects or reconciles two sources
6.5%
Enforces a procedure or checklist
6.5%
Something else
6.5%
What survey respondents built
What the app does
Share
Generates a report or summary
23.9%
Monitors conditions and flags issues
21.7%
Collects data from the field
13.0%
Calculates something specific to our equipment
13.0%
Handles a scheduling or assignment rule
8.7%
Connects or reconciles two sources
6.5%
Enforces a procedure or checklist
6.5%
Something else
6.5%
Reports and condition flags together are 45.7% of first builds. Q13, n=46, fielded as single-select.
For a long time, getting a custom tool meant writing up a request and waiting on IT, a
designer or a developer. Every handoff turned the request into a game of telephone. Now
the person who knows the workflow best can describe what they need and build it that
same day, and the tool they get matches what they had in mind.
People who have already built an app are much more interested in building more. Among
builders, 63.6% are very interested in building a custom app for their own work, compared
with 19.4% of people who have never used Nova.
Figure 13 Interest in building a custom app, by Nova experience
19.4%
Never used Nova
49%
Use Nova, have not built
63.6%
Have built an app
Interest in building a custom app, by Nova experience
Nova experience
Very interested
Never used Nova
19.4%
Use Nova, have not built
49.0%
Have built an app
63.6%
Among respondents who selected "I am very interested" in building a custom app. Correlation. Builders answered this question because of a survey branching error, which the report documents. Bases: 350, 49 and 44 respondents.
A simple way to get more of the crew building their own apps is to show them what
coworkers have already built. Among people who have never used Nova, 40.6% were unaware
building was an option or had trouble picturing what to make. A first build shared at a
shift meeting answers both, because the crew sees a working tool made by someone who does
their job.
Figure 14 What keeps people from a first build
Do not think I have the technical skills
22.9%
Did not know I could
20.9%
Do not have time to explore it
20.6%
Not sure what I would build
19.7%
Need approval first
7.1%
What we have already covers it
7.1%
Tried and it did not work
1.7%
What keeps people from a first build
Reason
Share
Do not think I have the technical skills
22.9%
Did not know I could
20.9%
Do not have time to explore it
20.6%
Not sure what I would build
19.7%
Need approval first
7.1%
What we have already covers it
7.1%
Tried and it did not work
1.7%
The top four sit within about three points of each other. "Did not know I could" and "not sure what I would build" together are 40.6%. Q13b, n=350.
Closing
How operations pull ahead in 2027
Plant A
Two plants with matching equipment and crew size both start 2027 with a supervisor
putting together the weekly downtime report by hand. By the end of the year, one
supervisor has built a tool that assembles the report automatically. The supervisor who
built the tool was the natural person for the job.
Each finished build frees up time for the next one, so the first plant keeps automating
work the second plant is still doing by hand.
Plant B
At the second plant, the supervisor keeps up with the report by staying late. Those
extra hours get logged as regular labor, so leadership sees a normal week and leaves the
process as it is.
The first plant got ahead because a leader gave the supervisor who was already doing the
report the time to build a tool for it.
Across this research, supervisors and managers make up 40.4% of everyone who has built a
working app, the most common role among builders by a wide margin. Supervisors and
managers lead maintenance teams day to day, so they see where a process slows the crew
down and have the expertise to fix it.
Figure 15 Who builds
Maintenance supervisors and managers
40.4%
Technicians and maintenance staff
17.3%
Other roles
17.3%
Reliability and engineering
7.7%
Directors and VPs
7.7%
Executives and owners
3.8%
IT
3.8%
EHS and compliance
1.9%
Who builds
Role
Share of builders
Maintenance supervisors and managers
40.4%
Technicians and maintenance staff
17.3%
Other roles
17.3%
Reliability and engineering
7.7%
Directors and VPs
7.7%
Executives and owners
3.8%
IT
3.8%
EHS and compliance
1.9%
Q2 × Q11, builders only, n=52.
Appendix
Methodology and full survey results
UpKeep fielded an online survey through SurveyMonkey from September 1 to September 14,
2026. The final sample is 518 maintenance and operations professionals across eight
industry groups and eight role groups. Respondents were drawn from UpKeep's customer and
prospect lists, so the sample reflects UpKeep users more than the market as a whole. About
78% work in maintenance or operations functions of 25 people or fewer, and about 10% are
Directors or VPs of Operations.
Team size
42.5% 6 to 25
35.3% 1 to 5
14.5% 26 to 100
7.7% More than 100
Team size
Team size
Share
6 to 25
42.5%
1 to 5
35.3%
26 to 100
14.5%
More than 100
7.7%
Industry
Manufacturing and industrial
29%
Facilities and property management
17.6%
Other
16.2%
Food and beverage
12.9%
Healthcare
8.5%
Education or government
8.3%
Fleet and transportation
4.1%
Energy and utilities
3.5%
Industry
Industry
Share
Manufacturing and industrial
29.0%
Facilities and property management
17.6%
Other
16.2%
Food and beverage
12.9%
Healthcare
8.5%
Education or government
8.3%
Fleet and transportation
4.1%
Energy and utilities
3.5%
Role
Maintenance supervisor or manager
32.8%
Other
25.9%
Technician or maintenance staff
18.5%
Director or VP of Operations
9.8%
Reliability or engineering
4.6%
EHS or compliance
3.5%
Executive or owner
2.9%
IT
1.9%
Role
Role
Share
Maintenance supervisor or manager
32.8%
Other
25.9%
Technician or maintenance staff
18.5%
Director or VP of Operations
9.8%
Reliability or engineering
4.6%
EHS or compliance
3.5%
Executive or owner
2.9%
IT
1.9%
Data notes
Correlation only.
Every relationship in this report is self-reported at a single point in time. Differences between groups describe association, and no finding establishes cause.
Q13 ran as single-select.
Each builder described one app and chose one category, so share of apps equals share of builders.
Field-period effect.
Enthusiasm and adoption measures declined over the field period, consistent with early responders being more engaged with the topic. All figures use the full 518.
Robustness checks.
A core-maintenance-roles cut, a low-effort response screen, and an industrial-only cut were each run. Headline figures held within about four points.
Small bases.
Director and VP figures (n=51), builder figures (n=44 to 52), and industry cells under 50 are directional.
Blank answers.
Q7 shares exclude respondents who skipped a row (bases 512 to 516).
Definitions.
"Paperwork" combines three Q9 answers: writing up work orders and documentation, finding information in manuals or past records, and reporting to leadership. "No AI deployed" combines two Q5 answers: not using AI and looking into it. "External pressure" combines four Q8 answers: competitors, customers and auditors, leadership mandate, and regulation.
Appendix
Full survey results
Every question in the survey, as counts and shares. These are plain
tables rather than charts, so the numbers can be read, copied and cited
without going through the PDF. Q5 and Q12 are not printed here because
the report does not publish their toplines; the Q5 shares appear in
Figure 1.
Q1Industryn=518
Manufacturing and industrial
29%
Facilities and property management
17.6%
Other (please specify)
16.2%
Food and beverage
12.9%
Healthcare
8.5%
Education or government
8.3%
Fleet and transportation
4.1%
Energy and utilities
3.5%
Response
Count
Share
Manufacturing and industrial
150
29.0%
Facilities and property management
91
17.6%
Other (please specify)
84
16.2%
Food and beverage
67
12.9%
Healthcare
44
8.5%
Education or government
43
8.3%
Fleet and transportation
21
4.1%
Energy and utilities
18
3.5%
Q2Rolen=518
Maintenance supervisor or manager
32.8%
Other (please specify)
25.9%
Technician or maintenance staff
18.5%
Director or VP of Operations
9.8%
Reliability or engineering
4.6%
EHS or compliance
3.5%
Executive or owner
2.9%
IT
1.9%
Response
Count
Share
Maintenance supervisor or manager
170
32.8%
Other (please specify)
134
25.9%
Technician or maintenance staff
96
18.5%
Director or VP of Operations
51
9.8%
Reliability or engineering
24
4.6%
EHS or compliance
18
3.5%
Executive or owner
15
2.9%
IT
10
1.9%
Q3Size of the maintenance or operations functionn=518
6 to 25
42.5%
1 to 5
35.3%
26 to 100
14.5%
More than 100
7.7%
Response
Count
Share
6 to 25
220
42.5%
1 to 5
183
35.3%
26 to 100
75
14.5%
More than 100
40
7.7%
Q4Personal AI use, at work or outside itn=518
I use them most days
37.1%
I use them occasionally
30.7%
I've tried them a few times
23%
I've never used one
9.3%
Response
Count
Share
I use them most days
192
37.1%
I use them occasionally
159
30.7%
I've tried them a few times
119
23.0%
I've never used one
48
9.3%
Q6How respondents feel about AI in their line of workn=518
I'm interested, but cautious
43.8%
I'm excited about what it could do
35.1%
I don't think it applies to work like mine
8.9%
I'm concerned about what it means for our team
6.6%
I'll believe it when I see it
5.6%
Response
Count
Share
I'm interested, but cautious
227
43.8%
I'm excited about what it could do
182
35.1%
I don't think it applies to work like mine
46
8.9%
I'm concerned about what it means for our team
34
6.6%
I'll believe it when I see it
29
5.6%
Q8Single biggest driver of AI adoptionn=518
We want to work more efficiently
42.9%
Nothing is driving it right now
19.5%
I'm personally curious about it
8.5%
Our competitors are already doing it
6.6%
Leadership told us to
6.2%
We're losing knowledge as people retire
5.4%
Other (please specify)
3.5%
Our tools can't do what we need
3.3%
Our customers and auditors are asking about it
1.9%
We have to cut costs
1.7%
Regulations require it
0.6%
Response
Count
Share
We want to work more efficiently
222
42.9%
Nothing is driving it right now
101
19.5%
I'm personally curious about it
44
8.5%
Our competitors are already doing it
34
6.6%
Leadership told us to
32
6.2%
We're losing knowledge as people retire
28
5.4%
Other (please specify)
18
3.5%
Our tools can't do what we need
17
3.3%
Our customers and auditors are asking about it
10
1.9%
We have to cut costs
9
1.7%
Regulations require it
3
0.6%
Q9Part of the work that would benefit most from AIn=518
Writing up work orders and documentation
23.2%
Planning and scheduling
17%
Finding information in manuals, records, or past work
14.9%
Diagnosing what's wrong with equipment
13.7%
None of the above
11%
Managing inventory and parts
7.7%
Compliance and audit preparation
5.4%
Reporting to leadership
5%
Training new technicians
2.1%
Response
Count
Share
Writing up work orders and documentation
120
23.2%
Planning and scheduling
88
17.0%
Finding information in manuals, records, or past work
77
14.9%
Diagnosing what's wrong with equipment
71
13.7%
None of the above
57
11.0%
Managing inventory and parts
40
7.7%
Compliance and audit preparation
28
5.4%
Reporting to leadership
26
5.0%
Training new technicians
11
2.1%
Q10How often the team changes its process to fit the softwaren=518
Sometimes
35.3%
Rarely
19.7%
Often
18.1%
Almost always. We adapt to the software.
15.8%
Almost never. The software adapts to us.
11%
Response
Count
Share
Sometimes
183
35.3%
Rarely
102
19.7%
Often
94
18.1%
Almost always. We adapt to the software.
82
15.8%
Almost never. The software adapts to us.
57
11.0%
Q11Nova app buildingn=518
No, I haven't used Nova yet
79.2%
No, but I use Nova for other things
10.8%
Yes, I've built one or more apps
10%
Response
Count
Share
No, I haven't used Nova yet
410
79.2%
No, but I use Nova for other things
56
10.8%
Yes, I've built one or more apps
52
10.0%
Q13What the app doesn=46, single-select as fielded
Generates a report or summary
23.9%
Monitors conditions and flags issues
21.7%
Collects data from the field
13%
Calculates something specific to our equipment
13%
Handles a scheduling or assignment rule
8.7%
Connects or reconciles two sources of information
6.5%
Something else (please specify)
6.5%
Enforces a procedure or checklist
6.5%
Response
Count
Share
Generates a report or summary
11
23.9%
Monitors conditions and flags issues
10
21.7%
Collects data from the field
6
13.0%
Calculates something specific to our equipment
6
13.0%
Handles a scheduling or assignment rule
4
8.7%
Connects or reconciles two sources of information
3
6.5%
Something else (please specify)
3
6.5%
Enforces a procedure or checklist
3
6.5%
Q14Main reason for building instead of using an existing featuren=46
I was curious what Nova could do
21.7%
I wanted it to work a certain way
19.6%
The workflow was specific to our facility
15.2%
The logic was specific to our equipment
13%
We were tracking it in a spreadsheet
10.9%
It was faster to build than to find
10.9%
Someone on my team asked for it
6.5%
Nothing on the market does this
2.2%
Response
Count
Share
I was curious what Nova could do
10
21.7%
I wanted it to work a certain way
9
19.6%
The workflow was specific to our facility
7
15.2%
The logic was specific to our equipment
6
13.0%
We were tracking it in a spreadsheet
5
10.9%
It was faster to build than to find
5
10.9%
Someone on my team asked for it
3
6.5%
Nothing on the market does this
1
2.2%
Q15What builders would have done without buildingn=46
I'd have kept doing it manually
50%
I'd have submitted a feature request and waited
13%
I'd have bought another piece of software
13%
I'd have hired an outside developer
10.9%
It never would have gotten built
8.7%
I'd have filed an IT ticket
4.3%
Response
Count
Share
I'd have kept doing it manually
23
50.0%
I'd have submitted a feature request and waited
6
13.0%
I'd have bought another piece of software
6
13.0%
I'd have hired an outside developer
5
10.9%
It never would have gotten built
4
8.7%
I'd have filed an IT ticket
2
4.3%
Q13bWhat has kept respondents from building an appn=350, respondents who have never used Nova
I don't think I have the technical skills
22.9%
I didn't know I could
20.9%
I don't have time to explore it
20.6%
I'm not sure what I'd build
19.7%
I need approval first
7.1%
What we have already covers it
7.1%
I tried and it didn't work the way I wanted
1.7%
Response
Count
Share
I don't think I have the technical skills
80
22.9%
I didn't know I could
73
20.9%
I don't have time to explore it
72
20.6%
I'm not sure what I'd build
69
19.7%
I need approval first
25
7.1%
What we have already covers it
25
7.1%
I tried and it didn't work the way I wanted
6
1.7%
Q14bInterest in building a custom app for their own workflown=350, respondents who have never used Nova
Neutral
30.6%
Somewhat interested
25.1%
Very interested
19.4%
Not at all interested
12.6%
Not so interested
12.3%
Response
Count
Share
Neutral
107
30.6%
Somewhat interested
88
25.1%
Very interested
68
19.4%
Not at all interested
44
12.6%
Not so interested
43
12.3%
Cross-tabs cited in this report
Change in the past 12 monthsQ7a to Q7e, blank answers excluded
Pressure
n
Decreased
No change
Increased somewhat
Increased a lot
Leadership pressure
516
9.5%
57.0%
24.4%
9.1%
Customer or auditor expectations
513
7.0%
72.9%
14.8%
5.3%
Competitor awareness
515
4.6%
55.1%
29.5%
10.7%
Budget for AI or automation
513
4.8%
67.6%
20.5%
7.0%
Falling-behind concern
512
6.4%
63.9%
23.2%
6.4%
Excitement by personal AI useQ6 × Q4
Personal AI use
n
Excited
I use them most days
192
67.2%
I use them occasionally
159
26.4%
I've tried them a few times
119
7.6%
I've never used one
48
4.2%
Paperwork against diagnosis, by roleQ9 × Q2
Role
n
Paperwork
Diagnosis
Maintenance supervisor or manager
170
44.1%
15.3%
Other (please specify)
134
41.0%
6.7%
Technician or maintenance staff
96
43.8%
19.8%
Director or VP of Operations
51
47.1%
11.8%
Reliability or engineering
24
41.7%
25.0%
EHS or compliance
18
44.4%
11.1%
Executive or owner
15
26.7%
13.3%
IT
10
50.0%
10.0%
Paperwork against diagnosis, by industryQ9 × Q1
Industry
n
Paperwork
Diagnosis
Manufacturing and industrial
150
42.0%
19.3%
Facilities and property management
91
50.5%
11.0%
Other (please specify)
84
33.3%
9.5%
Food and beverage
67
41.8%
14.9%
Healthcare
44
52.3%
9.1%
Education or government
43
39.5%
9.3%
Fleet and transportation
21
52.4%
9.5%
Energy and utilities
18
38.9%
22.2%
Deployment and concern by roleQ5 × Q2 and Q7e × Q2
Role
n
Org uses AI regularly
Concern increased
Maintenance supervisor or manager
170
34.7%
29.8%
Other (please specify)
134
19.4%
20.1%
Technician or maintenance staff
96
32.3%
20.2%
Director or VP of Operations
51
13.7%
56.0%
Reliability or engineering
24
25.0%
47.8%
EHS or compliance
18
22.2%
38.9%
Executive or owner
15
26.7%
40.0%
IT
10
50.0%
40.0%
All respondents
518
27.4%
29.7%
Builder rate by personal AI useQ11 × Q4
Personal AI use
n
Built an app
I use them most days
192
18.2%
I use them occasionally
159
6.3%
I've tried them a few times
119
5.0%
I've never used one
48
2.1%
Builder rate by industryQ11 × Q1
Industry
n
Built an app
Manufacturing and industrial
150
12.7%
Facilities and property management
91
11.0%
Other (please specify)
84
7.1%
Food and beverage
67
14.9%
Healthcare
44
2.3%
Education or government
43
7.0%
Fleet and transportation
21
4.8%
Energy and utilities
18
11.1%
Interest in building by Nova experienceQ14b × Q11
Nova experience
n
Very interested
Somewhat interested
No, I haven't used Nova yet
350
19.4%
25.1%
No, but I use Nova for other things
49
49.0%
24.5%
Yes, I've built one or more apps
44
63.6%
18.2%
Builder profile by roleQ2, builders only
Role
Count
Share
Maintenance supervisor or manager
21
40.4%
Other (please specify)
9
17.3%
Technician or maintenance staff
9
17.3%
Reliability or engineering
4
7.7%
Director or VP of Operations
4
7.7%
Executive or owner
2
3.8%
IT
2
3.8%
EHS or compliance
1
1.9%
Builder profile by industryQ1, builders only
Industry
Count
Share
Manufacturing and industrial
19
36.5%
Facilities and property management
10
19.2%
Food and beverage
10
19.2%
Other (please specify)
6
11.5%
Education or government
3
5.8%
Energy and utilities
2
3.8%
Healthcare
1
1.9%
Fleet and transportation
1
1.9%
Builder profile by team sizeQ3, builders only
Team size
Count
Share
6 to 25
27
51.9%
1 to 5
11
21.2%
26 to 100
8
15.4%
More than 100
6
11.5%
About the research
Questions about this survey
How the sample was built, what the bases mean, and what you can do with the numbers.
Who was surveyed, and when?
UpKeep fielded an online survey through SurveyMonkey from September 1 to September 14, 2026. The final sample is 518 maintenance and operations professionals across eight industry groups and eight role groups, from technicians through to executives.
Does the sample represent the maintenance market as a whole?
No. Respondents were drawn from UpKeep's customer and prospect lists, so the results describe UpKeep users more closely than they describe the market.
The sample also leans small. About 78% work in a maintenance or operations function of 25 people or fewer, and about 10% are Directors or VPs of Operations.
Can I cite these figures?
Yes. The data is published under a Creative Commons Attribution 4.0 license. Cite it as UpKeep, 2027 State of Maintenance: AI in Operations, with a link to this page.
Every figure on this page has a table view, and the full topline and cross-tabs are in Full Data above.
Do these findings show that AI caused the results described?
No. Every relationship in this report is self-reported at a single point in time. Where two measures move together the report says so, and no finding here establishes cause.
Why does the base change from one figure to the next?
Questions that were only asked of some respondents carry their own base. The app-building questions were put to builders, so they run between n=44 and n=52, and Director and VP figures are n=51.
Q7 shares exclude anyone who skipped a row, which puts those bases between 512 and 516. Industry cells under 50 are directional.
How were the headline figures checked?
Three cuts were run against the full sample: core maintenance roles only, a screen for low-effort responses, and an industrial-only cut. Headline figures held within about four points on all three.
Enthusiasm and adoption measures declined across the field period, which is consistent with early responders being more engaged with the topic. All figures use the full 518.
What do 'paperwork', 'no AI deployed' and 'external pressure' mean here?
Each one combines answers to a single question. Paperwork is three Q9 answers: writing up work orders and documentation, finding information in manuals or past records, and reporting to leadership.
No AI deployed is two Q5 answers: not using AI, and looking into it. External pressure is four Q8 answers: competitors, customers and auditors, a leadership mandate, and regulation.
What counts as having built an app?
Q11 asked whether the respondent had built a working app in Nova. Builders then described one app and chose one category for it in Q13, which ran as single select, so share of apps equals share of builders.
