The AI conversation on a Calgary show floor sounded different this fall.
At BUILDEX Alberta 2026, held in Calgary in late October, the technology stage drew the usual crowd. What changed was who was asking questions, and what they asked. According to Construction Owners' Day 1 report, superintendents, estimators and project managers wanted to know about setup time, whether a tool works with the scheduling and accounting software they already run, and who owns the project data.
The site's correspondent put it in three lines: Does it save my superintendent an hour a day? Does it talk to the systems I already use? Who owns my data?
Those are the right questions. The third one is the one most vendors would rather you didn't ask, and it's the one that matters most once a tool is inside your business.
Where these questions came from, and their limits
One trade-show report isn't a survey. It's one correspondent's read of one day of sessions, published by a site that covers the construction market. It tells you what people on the floor were asking, not how many of them.
It lines up with better data, though. A spring 2026 national survey by SiteNews and Procore found only 10% of Canadian construction respondents report widespread AI use across multiple workflows, and named fragmented project data, spread across disconnected systems, as the structural problem underneath. Procore sells construction software, so it has an interest in that finding. The point still holds: the questions Calgary crews asked out loud are the same ones the national numbers keep circling.
Procore's summary of the same survey adds detail that bears on all three questions. It says 87% of firms rely on general productivity AI, while only 14% use tools built for construction workflows like RFIs and scheduling, that 84% of organizations work in hybrid or fully siloed project environments, and that fewer than 5% of respondents name budget as their main barrier. Those figures come from the summary page, not the full report, and the summary doesn't state the sample size. Treat them as direction, not precision.
Statistics Canada's own figures point the same way. Construction sits near the bottom of Canada's AI adoption numbers, as covered in Construction Sits Near the Bottom of Canada's AI Adoption Numbers.
Why this matters more in Alberta right now
Alberta contractors aren't short of work. BuildForce Canada's 2026 to 2035 outlook for Alberta expects non-residential construction to keep rising across the decade, with light rail transit in Calgary and Edmonton, the Scotia Place arena and Calgary Arts Commons named among the projects driving it.
The same outlook says the province's construction labour force needs to add 48,800 workers by 2035, mostly to replace 43,700 expected retirements, and may still come up as many as 5,300 workers short.
That combination, more work and not enough experienced people, is exactly where AI gets pitched hardest. It's also where a bad decision costs the most, because the people who'd have to clean it up are the same superintendents and PMs who are already stretched. More on that in The AI Boom Is Reaching Contractors Through the Hiring Market.
The Alberta Construction Association has started treating this as a skills issue. In September it launched a Digital Skills Training Program with a non-technical course on AI in construction alongside BIM, reality capture and VDC. Training helps people use a tool well. It doesn't answer whether that tool should be in your business, or on what terms.
Question one: does it save my superintendent an hour a day?
This is the easiest one to get a confident answer to, and the easiest one to get wrong.
Every vendor has a time-saved figure. Most of them come from the vendor's best customers, under conditions you won't match. The only number that means anything is one measured in your own operation, on a task you've timed before the tool goes in.
That means picking one task, writing down how long it takes today and who does it, and agreeing on what "better" looks like before anything is installed. Daily reports, RFI logs, meeting minutes and submittal tracking are common candidates because they're repetitive, they're documented, and the person doing them can tell you quickly whether the output is usable.
If a vendor can't tell you which task their tool replaces, and how you'd measure it in 90 days, you don't have an hour-a-day tool. You have a demo.
Measurement is where a lot of firms fall down. The same survey summary says 43% of construction firms don't track the outcomes of their technology at all. A tool nobody measured is hard to defend at renewal time, and harder to expand with any confidence.
Question two: does it talk to the systems I already use?
Most Calgary GCs aren't starting from zero. There's an estimating tool, a scheduling tool, an accounting package, a document system, and usually a pile of spreadsheets holding the whole thing together.
A new AI tool that lives outside all of that creates a new problem: someone has to move information in and out of it by hand. That's double entry, and it eats the time the tool was supposed to save.
The national survey puts a rough size on that cost. Its summary describes manual double entry taking up to five hours a week from project managers. That's an upper figure, not an average, but it's a fair warning: a tool that adds one more place to type things can make that number worse, not better.
Before a demo, ask three things:
- What does it connect to, today? Not on the roadmap. Today, with your versions of the software.
- Who does the setup? Your people, the vendor, or a third party, and how many hours it takes before the first real job runs through it.
- What happens when it's wrong? Where does a person check the output before it reaches a schedule, a bid or an invoice?
That last question is where most of the risk sits. AI that drafts is useful. AI that pushes numbers into your accounting or pricing without a person reviewing them is a different decision, and it should be made on purpose. The same logic applies to estimating, covered in AI Construction Estimating: Where It Helps.
Question three: who owns my data?
This is the one to slow down on.
When you load drawings, specs, bids, cost history or daily logs into an AI tool, three things are worth knowing in writing:
- Who can use it. Can the vendor use your project data to train or improve its models, or share it with anyone else? Is that on by default?
- Where it goes. Where is it stored and processed, and which subcontractors or cloud services handle it along the way?
- How you get it back. If you cancel, can you export everything in a usable format, and is it deleted from their systems afterward?
Your bid history and cost data are some of the most valuable things your company owns. They're how you price work better than the next firm. Handing them to a tool without knowing those three answers is like leaving the estimating binder on the counter at the supplier.
If you wouldn't leave the estimating binder with a supplier, don't hand it to a tool before you've read the data terms.
There's also a legal floor. Alberta's Personal Information Protection Act governs how private-sector organizations in the province collect, use and disclose personal information. Project data often includes it: worker names on daily logs, contact details in RFIs, incident reports. Your contracts with owners may add their own confidentiality terms on top. A quick check with your lawyer before you upload a live project is cheap insurance.
None of this means avoiding AI. It means the data terms are part of the purchase, the same as price and support. That's the idea behind building governance in from the start, which is how we approach every workflow. More on that on the governance page.
What a good answer looks like
A tool worth considering usually has plain answers to all three questions:
- Time: a named task, a baseline you measured, and a check date.
- Fit: a working connection to what you already run, or an honest estimate of the manual steps that remain.
- Data: written terms on training use, storage, export and deletion that your lawyer has read.
It also helps to know what kind of tool you're looking at. A general chat assistant your people already use for emails is a different decision from a construction platform that connects to your drawings, schedule and cost data. The second can do more, and it touches more of what you'd hate to lose, so it deserves a harder look on all three answers.
If any of the three is vague, that's not a reason to walk away. It's a reason to run a small test on a closed project, with non-sensitive data, before anything touches live jobs or client information.
What to do with this
Before your next AI demo, write the three questions down and send them to the vendor ahead of time. The answers you get back in writing will tell you more than the demo will.
Then look inward. Which task in your operation costs a superintendent or PM the most time each week? Which systems would a new tool have to work with? Which data would you be comfortable putting into it, and which would you never let leave the building? Those three lists are most of a decision.
Working through those lists for one business, and finding where AI actually fits before any money goes to a tool, 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.
