The Edmonton Construction Association has a line in its course listings that sums up where a lot of contractors are this year.
Its AI for Construction Estimating course opens by saying AI is already changing how construction offices operate, and that "the question is not whether your team will use it" but "whether they will use it well."
That question has moved out of the office. The ECA now runs a separate AI for Field Operations course for superintendents, assistant superintendents, foremen, field engineers, safety coordinators and QA/QC staff. It covers daily reports, safety briefings, field issues and RFIs, look-ahead planning, quality control and trade coordination, and it promises to show crews how to protect confidential project information and keep professional judgment in the loop.
So the Edmonton question isn't really "should our field people use AI?" Some of them already do, on their phones. It's how to make sure they use it well.
Where this comes from, and the limits of the evidence
A course catalogue tells you what an association thinks its members will pay to learn. It doesn't tell you how many Edmonton contractors use AI on site, or whether the courses work. The courses are taught by an outside instructor, and the tool requirements (paid versions of Microsoft Copilot, Google Gemini, ChatGPT or Claude for the field course) reflect his approach, not a standard.
The volume is still telling. The ECA education calendar lists more than two dozen dates for AI courses between December 2026 and late 2027, covering workflows, estimating, project management, Bluebeam's AI tools and field operations. An association doesn't schedule that much of something nobody is asking for.
National data points the same way. A spring 2026 SiteNews and Procore survey, summarized on Procore's report page, says 87% of firms rely on general productivity AI, while only 14% use tools built for construction workflows like RFIs and scheduling. Procore sells construction software, so it has an interest in that gap, and the summary doesn't give a sample size. Read it as direction: most crews are reaching for a general chat tool, not something built for the job.
Why field teams are the right place to look in Edmonton
Alberta's construction market is busy, and a lot of the work is in the field. BuildForce Canada's 2026 to 2035 outlook for Alberta says engineering construction stays elevated into the medium term, with light rail transit in Calgary and Edmonton and the Yellowhead Mainline Natural Gas Pipeline among the projects sustaining it into 2029.
The same outlook expects 43,700 Alberta construction workers to retire by 2035, 21% of the 2025 labour force. Many of them are the experienced superintendents and foremen who write the daily reports, run the toolbox talks and know which sub is about to fall behind.
Field data is also where the daily pain is. At BUILDEX Alberta in October, the commercial project manager at Total Fencing Solutions, which runs about 40 workers in the field from Fort McMurray to Osoyoos on a busy day, told Construction Owners that tracking crew hours is the "biggest challenge every single day and every single week." That's a field-capture problem before it's an accounting one.
That's why field use matters. The paperwork a superintendent does at the end of a long day is exactly the kind of work AI can shorten. It's also the paperwork that ends up as evidence in a delay claim, a deficiency dispute or an incident investigation. Shortcuts there have a long tail. The wider hiring picture is covered in The AI Boom Is Reaching Contractors Through the Hiring Market.
Daily reports: the easiest win, with one rule
Most superintendents already have the raw material for a daily report: photos, a few voice notes, the crew count, the weather and whatever went sideways. AI is good at turning that pile into a clean, consistently formatted draft.
The rule is simple: the superintendent's name on the report means the superintendent read it. AI is confident, and it fills gaps. If a voice note says "concrete pour delayed," a tool may write a cause that nobody said. On a daily report, an invented reason for a delay is worse than no reason at all.
Two habits make this work:
- Dictate facts, not conclusions. Say what happened and when. Let the report record cause only where someone actually knows it.
- Keep the source. Save the photos and voice notes with the report, so anyone can check the draft against what was captured.
The national survey summary says manual double entry can take up to five hours a week from project managers. A daily report that's drafted once, from the field, and flows to the office without retyping is a direct answer to that.
Safety briefings: AI can draft, the crew still assesses
This is where "use it well" matters most.
Alberta's Occupational Health and Safety Code is specific. Under Part 2, an employer must assess a work site and identify existing and potential hazards before work begins, prepare a report of the results, repeat the assessment when a work process is introduced or changes, and involve affected workers in the hazard assessment and in controlling the hazards.
AI can help with the preparation. It can draft a toolbox talk on cold-weather concrete, turn last month's near-miss reports into a short list of themes, or put a hazard assessment template into plain language for a crew whose first language isn't English. Those are real time savings.
AI can draft the briefing. It can't stand on the deck and look at the edge protection.
What it can't do is the assessment itself. A hazard assessment reflects the site in front of the crew on that day, and the Code expects the workers who are exposed to take part. A generic AI-written hazard list, signed without a walk, is a document that looks compliant and isn't. Keep AI on the drafting side of that line. Why confident AI output still needs a human check is the subject of Don't Believe Everything the Robot Says.
Look-aheads, RFIs, quality and coordination
The rest of the ECA's field list sits in the middle ground, where AI drafts and a person decides.
Look-ahead schedules. AI can turn a master schedule and the week's notes into a draft three-week look-ahead. The superintendent still decides what's realistic, because the tool doesn't know the crane is booked elsewhere on Thursday.
Field RFIs. A photo, a marked-up drawing and a two-line description can become a clear, well-structured RFI in minutes. The question it asks still has to be the right one.
Trade coordination. Summaries of coordination meetings and lists of open items are easy wins. Quality control. Deficiency lists built from walk-through photos and notes are a natural fit. AI can group items by trade and location and draft the list for each sub. Someone still confirms each item on the walk, because a deficiency that isn't there costs goodwill and one that's missed costs money.
Commitments to subs, and anything that changes scope or money, go out under a person's name after they've read every line.
The pattern is the same everywhere: the tool removes typing, not responsibility. More on how construction teams are using AI day to day in Construction Firms Now Use AI Daily, and Not Just for Estimates.
What "using it well" looks like on a site
A short set of field rules covers most of it:
- Approved tools only. Pick the tool and the account your crews use, and check its data settings. Personal free accounts on personal phones are where project information leaks.
- What stays out. Pricing, contract terms, worker personal information and anything covered by an owner's confidentiality clause. Who controls that data once it's in a tool is its own question, covered in Who Owns Your Project Data?
- Read before you sign. Every report, briefing and RFI is reviewed by the person whose name is on it.
- Facts from the field. AI organizes what crews observed. It doesn't supply observations.
- Safety stays human. Hazard assessments are done on site with the crew, every time.
None of this needs a thick policy manual. It needs a one-page standard and a superintendent who models it. The ECA's office courses accept free accounts, which is fine for practice. Before field teams start using AI on live projects, the business should decide which tool and account they use, and say so in writing.
What to do with this
Start with one superintendent and one project. Ask them to time their end-of-day paperwork for a week: daily report, photos, any RFIs, the next day's plan. That's your baseline.
Then pick one task, usually the daily report, and run it through an approved tool for a month with the two habits above. Compare the time against the baseline, and ask the PM whether the reports got better, worse or just faster.
If the daily report works, add one more task, such as field RFIs or deficiency lists, and measure it the same way. Resist the urge to roll a tool out to every site at once. One superintendent who trusts it will do more for adoption than a company-wide memo.
Write down your field rules before the second project starts, not after something goes wrong. If your team needs training, the ECA's field and office courses are a local option. Training teaches people to use a tool. The rules decide where it's allowed.
Working out which field tasks are worth changing, what rules need to be in place, and how you'll know it's working, for one business, is what the AI Discovery and Readiness Assessment is for. It's a paid engagement with a defined scope: a discovery call, structured input from the people closest to the work, and a written report on where AI is relevant to your operation, what to do first, and what governance needs to be in place before you start.
