There’s a lot of noise about AI in construction. Most of it is either “this changes everything” or “it’s all hype.” The truth is more useful and more boring.
AI is good at some kinds of work and bad at others. Knowing which is which saves you from wasting money in both directions.
A simple test
AI tends to do well when a task is repetitive, text-based, and easy to check. It does badly when a task needs judgment about the real, physical, or legal situation, or when a mistake is expensive and hard to spot.
A simple test
Three signs a task suits AI
Where it helps
Writing and rewriting. First drafts of emails, bid cover letters, project updates to owners, and daily summaries. You edit, it saves the blank-page time.
Summarizing. Long email threads, meeting notes, and long documents. Useful when you need the gist fast. You still read the part that matters.
Sorting and routing. Incoming email and requests: what’s a quote request, what’s an invoice, what’s urgent. Somebody still has to act, but nobody has to dig through it.
Follow-up. Quotes that went quiet, reviews after a job, reminders for missing paperwork. Not hard work, just work that gets forgotten. Software that sends the nudge on schedule does what a busy person means to do.
Pulling information out of paperwork. Reading a form or document and putting the key fields into a system, with a person checking.
Capturing from the field. Turning a photo and a few spoken sentences into a written project note. A lot of useful information never gets written down because typing at a job site is annoying.
Where it helps most: turning an idea into a plan
Everything above is a task. The place I get the most out of AI is higher up, with the owner, the director, or the manager who has an idea and needs it to become work.
Here is how it goes for me. I brain dump an idea, loose and out of order, the way I’d say it to someone standing next to me. The AI sorts it and turns it into a goal. From the goal it builds a project. From the project it builds the tasks, and it drafts who each one belongs with, based on what it knows about the company and its people.
Launching a new initiative
From a brain dump to assigned tasks
This is where I get the most out of it. The idea goes in loose. A plan comes out.
- 1
Brain dump
The owner, director or manager says the idea out loud, messy and unsorted.
- 2
Goal
AI sorts the idea into a clear goal and what done looks like.
- 3
Project
The goal becomes a project, built from what it knows about your company.
- 4
Tasks
The project becomes tasks, each drafted for the right person.
- 5
You review
A person checks the plan before it goes out.
Normally that takes weeks. A new initiative needs a meeting to agree on the goal, another to split up the work, another to sort out who does what. In my experience, a lot of good ideas lose steam somewhere in that stretch. With the sorting done for you, you can have a first version of the plan the next morning, and the team starts from something real instead of a blank page.
Two things keep this honest:
- It only works as well as what it knows. If it doesn’t know your people, roles and how you run jobs, it will guess. Feeding it real company information is what makes the plan usable.
- A person still reviews it. The AI drafts the goal, the project and the assignments. The owner or manager reads them and corrects them before anything goes to the team. That review takes minutes, and it is where your judgment comes in.
It fits launching new initiatives especially well: a new service line, a new hiring push, a new safety program, a new way of handling bids. Any owner, director or manager can brain dump the idea and get the structure back.
Where it doesn’t (yet), or shouldn’t
Estimating on its own. It can help organize and sanity-check, but a cost number that goes out without an experienced estimator reviewing it is a risk. It doesn’t know your local labor, your suppliers, or that the site has a bad access road.
Contracts and legal language. It can summarize and flag things to look at, but it can confidently misread a clause. A lawyer or your own careful read has to be the final word.
Safety decisions. It shouldn’t decide whether a site is safe or a procedure is compliant. It can help you draft the toolbox talk. It can’t stand on the site.
Anything where it would be wrong and nobody would notice. This is the real danger. AI often sounds sure when it’s wrong. If there’s no check in the process, the mistake reaches the client.
Where to be careful
The danger is the mistake nobody sees
Two questions decide how much to trust it: how bad is a mistake, and how easily would someone catch it?
Client relationships. Automatic follow-ups are fine. But the call after something went wrong should be a person.
The short version
Where it helps. Where it doesn't.
Where it helps
- Writing and rewriting
- Summarizing
- Sorting and routing
- Follow-up
- Pulling data out of paperwork
- Capturing notes from the field
- Turning an idea into a plan
Where it doesn't (yet), or shouldn't
- Estimating on its own
- Contracts and legal language
- Safety decisions
- Anything where a mistake would go unnoticed
- The call after something went wrong
What this means in practice
In practice
Five habits that keep it useful
- 1
Start small
Something repetitive and easy to check.
- 2
Set the limits
What it may do alone, and what a person approves.
- 3
Review early
Check the first output closely, then decide how far to trust it.
- 4
Mind what goes in
Contracts, financials and personal data are a decision, not a default.
- 5
Measure
Not saving time after a month? Change it or drop it.
The honest summary
AI is good at taking the small, tedious pieces of the job off people’s plates, and at helping the people who run the company turn ideas into organized work faster. It is not a replacement for the experience that lets you read a site, a client, or a bid. Companies that get value from it tend to use it for the first and protect the second.
Want to talk through your own situation?
If you’d like to talk through where AI could help at your company, Book a free one-on-one marketing session. I’ll tell you honestly where I’d start, and where I wouldn’t.
Benjamin