Procore Is Closing the Loop: DroneDeploy, Digital Coworkers, and the Price Question
In under eight months, Procore bought AI agent company Datagrid, packaged its agents into Digital Coworker tiers, and agreed an $845 million cash deal for reality capture platform DroneDeploy. On our latest episode, Buro Happold CTO Alain Waha argued those moves add up to a compounding learning loop: a system that starts to machine-learn how construction physically gets built. Co-host Martin asked the question every buyer should be asking back: software keeps getting more expensive, so where is the cheaper outcome?
Procore has had a busy 2026. In January it acquired Datagrid, a vertical AI company built around agents for construction. In May it embedded Datagrid’s intelligence directly into the platform. On July 23 it launched three Digital Coworker packages. Six days later it agreed to acquire DroneDeploy for roughly $845 million in cash.
Individually, each move reads as product news. On our latest episode of Bricks, Bucks and Bytes, Alain Waha made the case that together they are something more interesting, and Martin countered with the question that decides whether any of it matters to the people paying the invoices.
Both halves deserve a proper look.
The Pieces Procore Just Put Together
The record, the eyes, and the hands
Start with what each piece does. Procore’s core platform is construction’s system of record: the drawings, RFIs, submittals and daily logs that describe what is supposed to happen. DroneDeploy is the eyes: drones, ground robots and cameras that capture what is physically there, used on more than 3 million jobsites in over 180 countries. The acquisition was agreed on July 29 and is expected to close later this year, pending regulatory approval.
The Digital Coworkers are the hands. Launched on July 23, the packages bundle Procore’s AI agents into three tiers: a Starter Pack with five ready-to-use agents covering Deep Search, Submittal Review, RFI, Daily Log and Contract Review; a Pro tier opening up a library of 20 agents; and an Enterprise tier that adds Agent Studio for building custom agents. An upcoming capability called Procore Skills will let firms teach the AI their own procedures and standards. Under the hood, the agents are powered by the Datagrid intelligence Procore acquired in January and embedded in May, reasoning across drawings, specs and photos.
Record, eyes, hands. The obvious question is what happens when you connect them.
Closing the Loop on How Construction Happens
Send a signal, watch the result, learn, repeat
Alain’s answer borrowed an idea he attributes to Stanford professor Martin Fischer: real progress comes from learning systems sitting on persistent platforms. Today, construction barely learns. A design goes out, a subcontractor asks questions, things get built, and most of what actually happened evaporates.
Now run the same project through the combined stack. The design lives in Procore. The subcontractor’s queries live in Procore. The drone flies the site and shows what was physically built, when, and how it differs from the plan. The responses go back through the same system. For the first time, Alain argued, a platform can observe the full cycle, signal out, result in, and start machine-learning the construction process itself, the physical reality rather than just the paperwork. Every project makes the model smarter, which makes the platform stickier, which attracts more projects. That is the compounding loop.
He was careful to frame it as an opportunity Procore now has rather than a finished capability, and Owen pushed on the same point: the learning has to come from somewhere, whether that is an internal AI team, more acquisitions, or both. Autodesk, for its part, is running a parallel play across the asset lifecycle, agreeing in May to pay roughly $3.6 billion for maintenance platform MaintainX.
Every consolidation move, decoded weekly
We break down the deals, the data plays and what they mean for the people actually buying the software.
Join 3000+ Readers
Agents Do Tasks, Humans Keep the Job
And the price question nobody has answered
On the agents themselves, Alain was positive and precise. He called the Digital Coworkers genuinely deployable, then drew the line that matters: the distinction between the task and the job. An agent can automate a task, reviewing a submittal, drafting an RFI response, searching the archive. The human keeps the job, because the job carries accountability, and as he put it, human in the loop really means trust. Silicon holds no professional indemnity insurance and it hallucinates. The agent’s role is to make the human’s outcome quicker, faster, better, deeper and more sophisticated, in his words, rather than to replace the person who answers for it.
Then Martin brought the conversation down to the invoice. Construction software subscriptions keep climbing, Procore’s own pricing has long been a sore point with customers, and now AI packages arrive on top. His challenge to the whole category deserves quoting in full.
I would like to see purposely built AI which is decreasing the cost of the outcome for the buyer significantly, rather than seeing increased, inflated prices of software packages.Martin, co-host of Bricks, Bucks and Bytes
That is the test. If Digital Coworkers cut the real cost of delivering a project, the loop is worth every token. If they simply add a line to the software bill, the industry has bought a story. If you run these agents day to day, we genuinely want to hear which one it is.
The Loop, Piece by Piece
What each part captures and what it teaches
| Piece | What it captures | What it teaches the system |
|---|---|---|
| Procore platform | Drawings, RFIs, submittals, daily logs, the intended plan | What was supposed to happen, and every question raised along the way |
| DroneDeploy | Aerial and ground reality capture from 3M+ jobsites | What was physically built, when, and how it differs from the plan |
| Digital Coworker agents | Task execution: reviews, searches, logs, contract checks | How work actually gets processed, and what good output looks like |
| Procore Skills (upcoming) | A company’s own procedures and standards | How each specific firm wants its work done |
Frequently Asked Questions
To add reality capture to its platform: DroneDeploy’s drones, robots and cameras provide the live visual record of jobsites that Procore’s document-based system lacked. Procore CEO Ajei Gopal tied the deal directly to delivering digital coworkers that track a project, make sense of it, and act. (Source)
Three tiers announced July 23: a Starter Pack with five ready-to-use agents (Deep Search, Submittal Review, RFI, Daily Log, Contract Review), a Pro tier with a library of 20 agents, and an Enterprise tier adding Agent Studio for custom agents. A capability called Procore Skills, which teaches agents a company’s own procedures, is coming. (Source)
Datagrid, acquired in January 2026, brought the multimodal intelligence layer now embedded in Procore AI. It lets agents reason across drawings, specs and photos while keeping Procore’s permissions and security, and it powers the agentic Actions and Triggers behind the Digital Coworkers. (Source)
Connecting the plan, the physical result and the responses in one system, so the platform can observe what was intended, what was built and what changed, then learn from the difference. Alain Waha’s argument, channeling Stanford’s Martin Fischer, is that this turns construction from a process that forgets into a system that compounds.
An agent automates a task, such as reviewing one submittal. A human keeps the job, because the job carries accountability, trust and professional indemnity, none of which software can hold. In Alain’s framing, the agent exists to make the accountable human’s outcome faster, deeper and more sophisticated.
Not yet. Procore entered a definitive agreement on July 29, 2026 to acquire DroneDeploy for approximately $845 million in cash, and the transaction is expected to close later this year subject to regulatory approvals and customary closing conditions. (Source)
Related Articles
Procore Just Paid $845M for Eyes, Not Software
Autodesk Just Spent $3.6B to Own What Happens After the Building Is Done
The Wild World of Pricing Models
