A new system can work perfectly and still sit unused. In construction, that usually happens on site first, where a supervisor has a way of doing things that already gets the job done.
This is how we approach AI change management in construction: who to involve before anything is built, how to make it work on a phone, and how to tell whether people are using it.
Why do new systems stall on site even when the software works?
Because the habit is stronger than the tool. A supervisor who keeps a list in the ute will keep doing it if the new system asks for more effort than the old way. The system can be sound and still go unused.
The pattern we see is consistent. The team with someone inside who owns the system takes it on quickly. The team without one stalls, even with the same software. We now line up that person before we build.
The office has its own version of this. Accounts may keep its own spreadsheet because it has never seen the new report match the old one.
That is why we look for quiet workarounds early. A list in the ute, a personal spreadsheet and a text message to the office all tell you what the system will need to replace.
A system without an owner has nobody to ask when something looks wrong, so people stop asking and go back to what they know.
Who needs to be involved before anything is built?
Supervisors, leading hands, admin, accounts and the estimator, plus one person inside the business who owns the system. That person answers questions, spots problems early and keeps the team using it after the builders have gone. Without that person, nobody in the business is left to ask.
Pick the owner early. Choosing someone after the build leaves them with a system they had no say in.
Choose someone the site respects and the office trusts, with the authority to ask for changes. If nobody has the time to test it, it will not stick. Testing is part of the owner’s job, so give them room for it.
The owner’s job in practice is to answer the first questions from the team, collect what is not working, and pass it back to whoever is building. If that person lacks the time or the standing to ask for changes, the builders will not hear about problems until the team has stopped using the system.
Everyone else gets asked what they do today, what they would skip, and what they would not give up. Those answers shape the build.
How do you make it work for a supervisor on a phone?
Fewer fields and fewer taps. Dockets, daily reports and pre-starts should be filled in on the phone and land against the right job straight away. If the new way is slower than the paper docket, supervisors will not use it, and they will be right.
Think about where a supervisor actually is when they use it: on a site, with gloves on, in sun or rain, often with a poor signal. Anything that needs a long form, a login each time or a reliable connection will be skipped. The test is whether it works on a bad day.
Test it with a supervisor on a real job, standing up, with a phone in one hand. Watch which fields they skip. Then remove those fields or fill them in for them.
Nothing should be retyped in the office. If a docket arrives on the phone and someone rekeys it, you have moved the problem rather than solved it.
What does training look like for site crews compared with the office?
Separate sessions, run on real jobs, with the people who will use the system. Site crews need to know what to tap and what happens next. The office needs to know where the information lands and how to check it.
Go-live is for using the system, not for finding out how it works. By then everyone should have already seen their own part of it on a job they know.
Short sessions near the job work better than a long one in a meeting room. Use real dockets and real job names, so the first time someone sees the system it already looks like their work. Give each person the one or two things they will do daily, and leave the rest until they ask for it.
We cover how we run this in how we work.
What do you do about the person who prefers the old way?
Ask what the old way gives them. They are usually right about something: a check the new system skips, a field a supervisor needs, a report only they read. Fix that, and the person who prefers the old way can become the one who explains the new one.
Listen for the reason before the objection. “I like my list” often means “your system doesn’t show me tomorrow’s jobs.” That is a build problem you can fix.
Give them a role in the rollout. Ask them to test the new system against their old checks and tell you where it falls short. People who help find the gaps tend to defend the result.
How do you know people are actually using it?
Look for plain signs, not a dashboard. Dockets land on the right job the same day. Variations get logged on site. Nobody is running a side spreadsheet. If any of those are missing, the system is not the only one in use.
A dashboard can look healthy while a spreadsheet runs in parallel. Walk to the office and ask what people open first in the morning.
Ask the supervisors as well as the owner. Ask the person who closes the month what they trust. Their answers tell you more than a usage report, and they show you where to put the next round of fixes.
Running old and new side by side is a separate topic, and we cover it in AI implementation for construction companies. If you are weighing up a rollout, you can also see results from jobs we have delivered.
