Procore AI went out globally last month. I've been circling this one for a fair while, asking questions, keeping across the announcements, waiting for the day it became something you could actually get. So I'm pleased it's here, and I want to say up front this is a positive development for anyone running Procore. It will ultimately make the platform better for you.
Because it's a first release, though, I think the more useful thing I can do is share how I'd think it through, the same way I'd talk it through with any of the teams I work with.
Some context first, Procore bought an AI company called DataGrid late last year, and rather than launching under yet another product name, they've kept it simple. DataGrid now sits at the heart of Procore AI, acting as the intelligence layer underneath it. On its own, DataGrid is a construction-focused AI platform that connects across a whole range of systems the industry uses. Inside Procore, that engine is pointed at your Procore data, your drawings, RFIs, submittals, daily logs, all of it.
Procore lists 18-plus specialised agents available out of the box, with the option to build custom ones on top. Most are fairly focused, one reviews submittals against your documentation, one drafts RFIs, one answers questions across your specs and drawings with citations back to the source. Each tends to do its one job and that focus is a deliberate design choice.
One practical note, you get to decide which agents are introduced into your account, it depends on your specific needs inside your business, have this conversation with your Procore consultant for more detailed information.
Whether it does what your team needs day to day is the first question, and I don't think anyone can answer it from the outside yet. Even sitting in the industry and working with Procore accounts, I'd want to see it applied in a real construction business, on live jobs, over a stretch of time. Procore has shared some early results from companies in their testing programs, and they're encouraging, significant time savings on document-heavy work. Whether that's realised at scale across the industry is what the next six to twelve months will show us.
This section applies to every AI product on the market, it's the same thing no matter the AI logo. Once you move to a paid plan, there are two sides to weigh, the dollar side, and the usage side, the tokens.
The tokens are the side I think will surprise people. I was speaking with some who described this as an entire whole area, tokenomics, there's a growing little niche of people just trying to work out how token usage translates into real-world cost for each industry. Most people's experience with AI so far has been a free plan or a modest monthly one, used here and there out of curiosity. Pointing it at real project work is a different scale of usage. You can be fully paid up and still work through your credits. I'm on a top-level Claude plan for Knight Solutions, and by Thursday afternoon I'm checking my usage. If Procore AI does meet your expectations and goes off like a frog in a sock with your live projects, getting things done left, right and centre, then usage is the thing to keep an eye on.
This is the question I keep coming back to, and it's an industry-wide one rather than a Procore one. Over the past couple of years just about every software company has built AI into its own product, Procore, ClickUp, HubSpot you name it, pretty much all of them. Each one is deep in its own platform, and depth matters. If I'm doing a detailed review inside the drawings tool, AI trained on Procore and sitting on my Procore data is exactly the right thing.
Thought my day doesn't stay inside one platform, and I'd say yours doesn't either. Say I need to review a set of drawings, check them against Australian standards, look back at a past project, and compare quotes from last time, now I'm reaching across several systems at once. So the question becomes, where does AI inside one platform, fit alongside the rest of what you run? The broader market will keep consolidating, it always does with these industry shifts, and you end up with a handful of major players. Where that lands, I couldn't tell you, and I'm wary of anyone who claims they can.
Get your hands dirty. DataGrid itself has a free plan available, you can sign up, connect a couple of things, and get a feel for how this style of AI works, outside of Procore, before any commitment. If you've barely touched AI at all, start even simpler, a free AI chat account, or a month of a paid plan, just to understand what it is and what it isn't. The old line of rubbish in, rubbish out has never applied more than it does here.
And that's the deeper point. AI is the ultimate amplifier, it will amplify good and it will amplify bad. Icing on the cake won't make a bad cake taste better. If the fundamentals of how your business runs its projects aren't sorted, AI will amplify that too. Getting the foundations right matters more now than it ever did with automations and workflows.
From my own experiences, here's where I've landed with AI in my own business, it's a sidekick. I hand it the grunt work and it does that work almost as well as I would, sometimes better. It never does my job start to finish, I'm constantly correcting it, redirecting it, catching things it missed. While it's smarter than me in places, that doesn't make it better.
If you've got Procore and you're curious, talk to your Procore consultant and request a demo, you want to see it against your own projects, with your own data.
For now, that's how I'm seeing it. Excited it's here, and asking a few questions rather than treating "it's AI and it's in Procore" as the whole assessment. Will it do what your team needs? Does the value stack up once you've weighed both sides of the cost? And where does it fit among the systems you already run? Those are yours to work through. I'll be watching the same things you are.