Construction software firm Procore Technologies, Inc. today announced plans to release a set of artificial intelligence agents early next year that are tuned to enforcing project efficiency, improving safety, enhancing decision-making and streamlining workflows.

Agents are based on Procore AI, a unified platform that combines generative AI with predictive analytics. They will aim to automate routine tasks, reduce manual workloads and surface insights that make construction projects faster and more efficient.

Among the specific tasks agents will tackle are complex processes such as managing requests for information, scheduling and submittals, which are documents or physical samples that a contractor provides to the project owner, architect or engineer for approval. RFIs can take days to complete and hundreds may be involved in large construction projects. Procore figures agents can cut completion times to hours.

An Agent Studio set to be released next year will enable users to customize agents to meet their unique project requirements without any coding, the company said.

Procore, which reported more than $1.1 billion in revenue in its most recent fiscal year, “has a 10-year history of using AI in construction,” said Chief Product Officer Wyatt Jenkins. “We use natural language processing to extract information from documents and have been doing computer vision for a long time to track progress and safety issues on construction sites.”

Construction has typically been a low-tech market dominated by paper, but the business is right for advanced technology like digital twins, virtual reality, predictive planning and simulations.

Grand View Research Inc. estimates the global construction and design software market will be $11 billion this year and grow more than 10% annually through 2030.

Beyond bots and drones

Procore already uses data capture techniques such as robots and drones to track construction progress and flag safety issues. “We have millions and millions of photos and videos of job sites,” Jenkins said. “We’ve got a rich history of computer vision, predictive analytics and natural language processing used throughout the platform.”

One popular feature uses computer vision to examine blueprints and computer models to figure out the exact number of parts that will be needed. Another scans images and videos to identify safety risks.

Large language models have kicked the use of AI up a notch in the industry, Jenkins said. “One of the most important things we do in construction is help people find answers to questions,” he said. “If you walk a job site, you’ll hear people asking about what’s in the contract and how long a footing should be in place. These are simple questions but the answer is buried in a construction document somewhere.”

Procore introduced a copilot last year and expects agentic AI to take functionality to the next level. One popular agent in early testing notifies project managers are hidden scheduling risks. “Construction’s all about time and money,” Jenkins said.

The combination of predictive analytics and LLMs will enable agents to analyze large volumes of text and data to generate schedules, predict anomalies and proactively notify managers about developments that merit their attention.

“An RFI agent can decide that a conversation should be headed in a particular direction and it’s not, so it gives recommendations to how we might resolve a problem based on other RFIs on similar projects,” he said.

The schedule agent can identify delays and automatically trigger notifications to affected contractors. “If there’s a schedule change that moves painting back a week, that might affect glazing, window dressing and other downstream trades,” Jenkins said. “The schedule agent will choose and notify all the parties involved.”

He said Procore is guarding against bias and hallucinations by citing source material in answers and applying human experts to validate responses.

Photo: Unsplash