Getting the Most Out of AI Takes Real Work, Say ENR FutureTech Speakers

AI was, again, a hot topic among the more than 700 construction professionals, architects and engineers in attendance, but deep dives into robotics on construction sites, cloud-enabled workflows, 3D printing of structures and better relationships between all stakeholders also came through in the program. 

“95% of AI pilots fail to deliver value,” said Alan Espinoza in his opening keynote address. The founder of Reconstructive AI and a veteran construction technology expert with firms such as Jacobs and Universal Creative, Espinoza offered hard-won lessons on making new technologies work in project delivery. 

“Is this an implementation problem?” he asked. “AI doesn’t fix broken workflows. It, rather, amplifies them… In construction, everyone is acting in their own self-interest and destroying the ecosystem they are all working in. The prisoner’s dilemma says people will act in their self-interest if not incentivized not to.”

Espinoza added that many of the transitions from 2D CAD workflows to building information models brought along bad data and communication practices into these 3D, cloud-based workflows. “We digitized the mess [with point solutions],” and didn’t address holistically what designers and contractors were trying to accomplish, he said.

“A strategic framework is what you need to do to start an implementation plan,” for AI, he said. “Align incentives, have whiteboard sessions, do anything to make them want to do the work.”

And this work doesn’t have to be overly dry and dull. Espinoza noted that one such session involved a “Dungeons and Dragons”-themed workflow with a “coordination swamp” and stakeholders cast as warriors and wizards to help team realize that construction challenges can be broken down into the kinds of surmountable challenges faced by players in role-playing games.

On a panel considering the effectiveness of AI agents, David Letteer, director of artificial intelligence at Hensel Phelps, said agentic AI “is the new cloud,” and that adoption is the hardest part. Proper data governance must come first, he said, and the autonomy of the agents must be earned, like that of any new hire.

Enrico Bertucci, McCarthy Holdings vice president, said the firm was using the Glean agent builder and has already deployed about 25 agents to sites across the U.S. after standardizing a process to approve them before they go to work.

Getting Results from Reality Capture

Trevor Owen, director of reality capture at Rogers O’Brien, showed how its purpose-built reality-capture robots named Rosa and Mac are saving 10 to 60 hours a day once spent by workers taking photos of the contractors’ work. 

The four-wheeled, quadruped robots can take photographs and laser scans. The highly mobile robots can navigate stairs and recover on their own from trips or falls. Their upload and data-sharing process were optimized by Owen and the robot provider Alpha Z to create faster updates to 3D models and construction documents so it actually makes data sharing faster on Rogers-O’Brien sites. 

“By documenting site conditions in real time, Mac empowers project teams to make faster and more informed decisions,” Owen said.

3D Printing a Walmart

Zach Mannheimer, chairman and founder of Alquist 3D, said his his presentation that Alquist’s robotic arm-enabled 3D printing process saved Walmart over $100,000 on a Huntsville, Ala. Supercenter project, and the company has already signed a deal with the retail giant to print buildings across the U.S. While discussing the twists and turns of 3D printing of concrete making it into commercial construction after initially being developed for residential homes, he explained “we had to go commercial to make a project scale.” 

Mannheimer said the robotic arm system Alquist and its customers use today costs about one quarter of a gantry 3D printing system, originally thought to be the best way to print concrete.