Procore's AI Go app lets jobsite crews ask their project data out loud

Procore AI Go is a mobile app that lets construction crews ask questions out loud and hear spoken answers pulled from their own project data. The app is available now in open beta. Procore launched it on Oct. 21 at its Groundbreak 2026 conference in Orlando, along with a partnership with Meta to bring the same voice experience to AI glasses.

The pitch targets a cost every jobsite knows. According to Procore, productivity drops and decisions can slip whenever crew leads must pause the job to find drawings, check specifications or record what took place in the field. Geoff Lewis, Procore's SVP of Product, said construction crews should be able to get the information they need without a walk back to the trailer. Contractors will want to know two things: how much of that gap a voice layer on top of existing records closes today, and how much of the larger promise rests on features that are still on the roadmap.

What a foreman can ask on day one

AI Go connects to a customer's existing Procore project and answers questions about drawings, specifications, RFIs, submittals, daily logs, observations and inspections. Answers come back spoken. Procore says the app understands project context, so it can surface the relevant document with supporting evidence and handle follow-up questions without the user starting over.

Conversation history ships at launch too. Crews can go back to earlier sessions instead of starting from scratch, and that history is stored on their devices. That's a small feature, but it fits a crew lead who asks about a detail in the morning and comes back to it after lunch.

One design choice says a lot about who it's built for. On the phone, crews use push-to-talk rather than an always-listening trigger, which Procore says keeps jobsite noise and nearby conversations from setting it off.

Procore's AI product page explains what sits underneath. It describes a proprietary reasoning model called Magpie, trained on industry data to work out relationships across messy, real-world jobsite records rather than just summarize text. It also describes embedding models built to read construction formats such as drawings, blueprints, PDFs, audio and video. Procore says that base draws on more than 20 years of industry data and 3 million global builds. For a foreman, the test is practical: does a question about a detail bring back the right document and the evidence behind it?

Then there's trust. Procore's support documentation says Procore AI cites its sources and stages any proposed change for human review before acting. That statement describes Procore AI as a whole rather than the mobile app specifically. For AI Go itself, Procore says the app can surface the relevant information and supporting evidence.

The product page adds that permissions follow a customer's existing roles and that agents never act without sign-off. If that holds in AI Go, a foreman's voice question would reach only what their Procore role already allows, which matters to any contractor wary of handing project-wide search to every phone on site.

Push-to-talk on the phone, "Hey Procore" on the glasses

The Meta partnership pushes the voice idea one step further. Procore plans to bring AI Go to AI glasses, so crews can pull up project information and take first-person photos by saying "Hey Procore" instead of reaching for a phone. For now Procore calls it exploratory. The collaboration will test how glasses could extend the app hands-free on active jobsites, including getting information about the images crews capture.

The two interfaces work against each other. For the phone, Procore picked push-to-talk on purpose, so background noise and nearby conversation would not set off the app. For glasses, a spoken wake phrase is the main feature, since the worker's hands are occupied. Lewis said the aim is to let crews look up and record information while keeping their tools in hand. Testing on active jobsites is what will reveal how well a wake phrase works amid the noise of a busy site.

Voice on the jobsite isn't new at Groundbreak. A year earlier, at Groundbreak 2025, Benetics AI announced it was expanding across the U.S. and into Canada after deployments with contractors in Michigan. Benetics markets itself as a voice assistant for site reporting. Procore's version runs inside the platform where the RFIs, submittals and logs already live.

From reading the record to writing it

The roadmap is where the app gets more ambitious. According to Procore, crews will soon be able to use AI Go to build daily logs and observations, write draft RFIs and translate information just by talking to the app and sharing images. The goal is to convert what workers see and say on site into structured project information while the context remains fresh.

Some of the groundwork for that move from lookup to record creation is already in place outside the app. Procore's AI page says more than 150 built-in actions let agents create or update records such as RFIs, change events, daily logs and action plans inside the platform. Agents can also fire when a relevant project event happens or run on a recurring schedule. In practice, voice capture in AI Go would give record types Procore AI already knows how to write a new way in.

Agents follow after that. Procore plans to add the Daily Log Agent to the app. It pulls together activity notes, photos, labor data and site details, and it catches information that is missing or conflicting before anyone finalizes a report. The Site Safety Agent is built to help crews identify potential hazards, such as a possible fall risk, and follow up on them.

The Site Safety Agent arrived in July as part of Procore's expanded agent library. At the time, Procore said it identifies hazards from site photos, videos and drawings and returns field-ready findings. Put that next to the first-person photos Procore wants to capture through glasses, and the direction of the field roadmap looks clear: images taken at the work face feeding an agent that flags what needs follow-up.

Credits, packages and the field budget

Procore's support documentation says Procore AI is available with the purchase of a Digital Coworker package. The packages come in three tiers: Starter, Pro and Enterprise. Starter includes five agents: Deep Search, Submittal Review, RFI, Daily Log and Contract Review. Enterprise adds Agent Studio for building custom agents.

Pro sits in the middle. When Procore introduced the packages on July 23, it said Pro opens a library of 20 agents covering workflows from project planning and bidding to safety, quality and risk management. Agent Studio is exclusive to Enterprise.

Every tier includes Control Tower, which shows administrators AI usage and credit consumption by agent, project or team member. The support site also has admin tutorials for managing AI budgets. That matters once the agents reach the phone. Because each package comes with a different set of agents, a contractor's tier could shape what its field crews can actually do in AI Go once agents arrive.

The two agents named for the app show how. The Daily Log Agent is one of the Starter Pack's five. The Site Safety Agent isn't. It belongs to the larger library that Pro opens up. A Starter customer and a Pro customer could therefore end up with noticeably different field apps.

Skills could matter in the field too. Procore said it would start rolling Skills out across all Digital Coworker packages in August. Skills let companies teach Procore AI their own processes through plain-language prompts or documents such as standard operating procedures. Thiago da Costa, Procore's SVP of AI & Data, tied the feature to preserving institutional knowledge as experienced workers retire, change roles or leave. If field agents apply those Skills, a daily log drafted by voice would follow the company's own standards rather than a generic template.

For contractors already on a Digital Coworker package, the app could raise a rollout question: which crews get it first, and how their usage shows up in Control Tower. For firms still evaluating Procore AI, the mobile app could add a field use case to weigh in the decision.

What the productivity numbers measure

It helps to know what each of the figures cited around the launch actually counts.

The hours figure comes from a joint Procore and Dodge Construction report, and it covers technology adoption in general. The Haskell result, from Procore's July announcement, is a document review run against project data. It's a big gain, but it isn't a foreman asking a question at the work face. Neither number measures what AI Go does, so for now any field-time savings from voice lookup are a reasonable expectation, not a demonstrated result.

Haskell was one of the early adopters helping shape Procore AI, along with Consigli Construction and Level 10 Construction. Hamzah Shanbari, Haskell's director of innovation, described the goal as reducing "that cognitive load of trying to find the right information." AI Go goes after the same problem, just at the work face instead of the office. That overlap makes the Haskell number tempting to borrow, and it's also why it shouldn't be.

The early customer feedback is about how the app feels to use. Andrew Baker, CIO at Sequoyah Electric, said his foremen can ask a question while looking at the problem instead of heading back to search drawings and specs. He added that the app feels natural because crews already work through issues by calling teammates.

Baker said that once crews try it, "they want to keep using it." That kind of familiarity counts on crews that don't love new software. It's still one contractor's experience in a beta.

Procore is taking part in a study with Moss Construction, independently designed by SBK Technologies, that will look at how AI agents, AI Go and AI-enabled wearables could reduce administrative work for crews. Procore says the study is meant to better understand AI's impact on field productivity. When results come out, they would be the first figures here that measure AI Go itself rather than technology adoption or office document review.