Three Roadblocks to AI Adoption in Construction (and How to Overcome Them)

While barriers remain for construction to fully unlock the potential of artificial intelligence, there are solutions the industry can focus on to help foster adoption that drives value, according to Kaushal Diwan, leader of WND Ventures, the venture arm of DPR Construction. Diwan presented on the topic at the BuiltWorlds 2025 Venture East Conference in Boston.

“In construction, AI isn’t likely to replace people as much as be a tool leveraged by talented people that helps address challenges like improving safety, procurement and so many other things throughout the project lifecycle,” Diwan said. “There are still barriers to AI adoption, but there are some strategies that can help overcome them.”

It’s been an issue in the construction industry for years: data is often in too many places or systems and, often, the data collected is not detailed enough to work for large learning models.

“Safety is a great example here,” Diwan said. “Our industry has a ton of data around incident rates, but the ways incident are reported, and where that information is kept, is very inconsistent. With quality data, AI would be able to recognize patterns more easily, but with inconsistent data, you get inconsistent results.”

Diwan thinks that data standardization would be a good step but also points to the need for leadership support throughout any construction organization. Startups that recognize the complexities of data in the construction industry may also be positioned well to help.

“DPR has support for this throughout its various levels of leadership,” Diwan said. “We also have a steering committee of people from throughout our business who look to identify the right opportunities to recommend for pilot projects.”

Across industry, IT teams are already dealing with staying out front of sophisticated attacks. AI presents new ways for malicious individuals to try and get into proprietary data.

“Security teams are rightly cautious,” Diwan said, “But many examples of company policies are either too restrictive or nonexistent. Both extremes slow down experimentation. In the meantime, improperly deploying AI tools in open enterprise environments creates gaps for cybersecurity threats.”

To get over this barrier, Diwan believes security standards should be practical and simple; they should support the natural curiosity of employees and give them the ability to experiment.

“Work with your cybersecurity and IT teams to help not only protect the business but also come up with solutions on how AI can be deployed in a safe/secure environment,” Diwan said. “That will create more certainty for where AI can or can’t be deployed internally.”

Construction has a wide range of tech fluency. Face-to-face communication is often favored by operations staff vs. email. Different roles employ different tools and, while some workers use AI every day, others rarely use digital tools. Adoption requires empathy, patience and low-friction interfaces, whether at a desk or the jobsite.

“For the curious, it’s important to guide potential users to tools where it’s safe to experiment,” Diwan said. “DPR rolled out Microsoft Copilot at an enterprise level for all admin employees while also making enterprise licenses of ChatGPT available. That gave a lot of folks on the fence a way to lean in.”

Diwan noted there’s a cost to deploying AI at scale, but done well, it provides strategic advantages.

“There’s a race to see how the right tools can be applied to deliver more value for customers, keep people safe and accelerate schedules,” Diwan said. “Five years from now, AI will have changed a lot about the ways we build. The question isn’t if AI will transform construction—it’s how fast companies can adapt.”