Are you struggling with applying artificial intelligence (AI) to your construction business? You might do well to think of AI as a new “colleague” who needs onboarding just like any other employee, said Murillo Piazzi, digital consultant at the BIM Academy, on a recent 21cc podcast, "AI: Friend, Foe, or Confused Colleague?"
Think of AI as a colleague who just started working at your company, Piazzi said. You wouldn't throw them in the deep end on day one with a massive, nuanced project. Instead, you'd train them, help them understand your business, and give them the preparation required to help them make intelligent decisions or know when to ask questions and seek additional information.
To begin, use AI first on tasks that don’t require nuanced thinking and that have clear objectives and verifiable outcomes. As a tool, AI still generally cannot produce ethical, nuanced assessments of situations, Piazzi said.
As an example, AI “doesn’t have ethical considerations [and doesn’t understand] something will hurt someone” in a construction project if done improperly, he added. “Leave those kinds of projects to humans,” Piazzi said.
Looking at AI usage from another perspective, David Philp FCIOB, chief value officer at Cohesive Group and chair of CIOB's innovation panel, agreed that clearly defined objectives were the key to success. Still, he also stressed that organizations must make a closer appraisal of their own structure and decision-making process before flipping the on switch with AI. "It all starts with good quality data," he said, emphasizing that AI is not a "panacea" or band-aid that you can slap onto a problem to provide an automatic fix.
Finally, be warned that your new AI colleague can make mistakes, too, warned May Winfield, Buro Happold's global director of commercial, legal, and digital risks. While poor data input will generally undermine AI's work, AI itself can also "hallucinate" and make things up sometimes in an overly ambitious effort to fill in gaps, she said.
There are several contractual and legal frameworks to consider when looking to implement AI, Winfield said. Users must be aware of copyright and reliability issues, among other considerations, she said. “Don’t blindly take AI output,” Winfield stressed.