I've been having the same conversation a lot lately. Different clients, different sized operations, different parts of the country, but it's fundamentally the same question. Where is Procore going with AI?
And I kind of get why people are asking, because even paying close attention, the last 12 to 18 months have been confusing.
Procore started out working with Microsoft, using Copilot as the foundation for some AI features inside the platform. They released a few things. You could opt into a beta. Some of it was actually useful. But then they sort of pulled back, and that was that. Features that people had started using were effectively put on hold. If you'd invested time in learning those tools, you were kind of left in limbo.
Then late last year, they acquired a company called DataGrid, an AI and data platform that was already well established in the construction tech space. Along the way, the AI branding inside Procore changed at least three times. They landed on Helix for a while, which I don't think registered with most of the people I talk to. And if you tried to use AI to find out what Procore is doing with AI, which I did, just to test it, the information you get back is already out of date. It still references the Microsoft partnership. Doesn't include the DataGrid acquisition at all.
So yeah. I don't blame people for being confused.
If you tried to use AI to find out what Procore is doing with AI, the information you get back is already out of date.
It was already a fully built product before Procore bought it. A few hundred connectors to different systems, SharePoint, Teams, Airtable, and it was already heavily geared towards construction. It already connected to Procore before the acquisition even happened. So this wasn't a situation where they bought some early-stage technology and had to build everything from scratch. The plumbing was already there. Procore absorbed the entire DataGrid team as part of the deal, so the people who built it are the same people running it inside Procore now.
The contrast with the Microsoft approach is worth mentioning. With Copilot, Procore had to build all the infrastructure around a general-purpose AI platform that wasn't designed for construction. DataGrid was. So the speed of integration has been different.
Procore then released their first five agentic agents built on DataGrid that work inside the platform. That's four or five months after the acquisition. Faster than I expected, if I'm honest.
What this starts to change, and this is the part I've been spending time trying to work out, is how you interact with the software day to day. Right now, using Procore means logging in, navigating to the right project, finding the right tool, clicking through menus, building reports. With DataGrid's AI built in, you can ask a question in natural language and get an answer. You can drop a photo into the mobile app and say "create an RFI about this waterproofing membrane" and it creates it. You're not navigating the interface the same way. The interaction starts to go through the AI rather than through the screens and menus that were originally designed.
DataGrid connects to hundreds of systems beyond Procore. So it's not just AI inside one platform. It can pull information from SharePoint, from your email, from scheduling tools, and surface it alongside your Procore data. No construction operation I've ever worked with runs on a single system, there's always five or six things alongside each other, and until now they've been pretty siloed. This is sort of the first thing I've seen that could realistically bring most of that together without needing custom integrations for every connection.
The contracts administrator spending half their day chasing documents across three systems is the person whose job looks different because of this.
My old boss used to say that construction, wherever you go, is one bloke nailing two pieces of wood together. And he's right. What AI changes isn't the physical work on site. It's the admin, how you find information, how you create records, how you identify where the risks are sitting before they become problems.
There was a billboard on a building site, where a major builder wrapped their scaffold signage with the words "Hey ChatGPT, finish this building." It went kind of viral because it captured something. The AI can sort through your RFIs and compare drawings. It cannot pour concrete.
I keep saying the same thing to people who ask me about it, don't try to boil the ocean with this stuff. Start with the basics. Get comfortable with what the terms mean, how the technology actually works at a simple level, and where it connects to what you're already doing. You can't fully know what you can do with it until you properly understand the processes inside your business. And once you start asking the right questions, it kind of snowballs, you solve one problem, and you realise it can probably solve three more.
The thing I'd push on, though, and this is based on what I'm actually seeing, is that it needs to be coordinated. I've already come across situations where individual people on a project team have gone and got their own AI subscriptions, started feeding project data into them, and nobody above them knows it's happening. That's fine when it's one person. When it's 15 or 20, all using different tools, putting project information into systems with no governance around it, that's where you start running into problems around security, data sovereignty, and contractual obligations.
I'm still learning a lot of this myself, if I'm honest. Every few days something changes. There's a whole new set of language that's come in fast, MCPs, connectors, agents, tokens, and even keeping up with just the Procore side of it has been a fair bit of work.
But I think there's value in getting the basics down now. So that when the next round of changes comes through, and it will, you've got something to anchor it to. And based on what I'm seeing, that next round won't be long.