The old computer mantra "garbage in, garbage out" is all too relevant today for construction industry professionals hoping to maximize artificial intelligence in their projects.  

At a recent panel discussion, "AI in Construction: Your Questions Answered," part of the Digital Construction Summit on September 24, sponsored by Trimble, speakers were bullish about how AI can potentially and significantly enhance construction tasks. However, this can only happen if project managers are clear from the outset about what problem they want to solve when leveraging it while ensuring that the incoming data is vetted, sourced and understood.

The “transparency and trust questions [about data] always comes up,” Karoliina Torttila, director of AI at Trimble Concerns, said at the event.

"Any data you use has to be accurate, relevant, and trustworthy," said Vicky Reynolds, CTO at Catalyst, stressing that the task isn't easy.

Vetting incoming data will be "on a case-by-case basis. Unfortunately, there's no easy five-step plan to developing the best data set for your AI, but the great thing about it is [that] it is [based on] all of the same practicalities and processes around good project management,” added Reynolds.

Know at the outset why you are using AI and what you are looking to address, she stressed. "It comes back to identifying your problem and determining how you are going to use this AI tool to make life easier," she said.

And avoid falling into the trap of collecting too much data. Misplaced enthusiasm, or caution, sometimes leads to situations where a person is afraid to leave out any data. They have a vague sense it may be helpful someday, "or just in case," she said. "It's really about using the least amount of information to gain the best possible outcomes," Reynolds said.  

Effectively sourcing data requires some detective work, determining where it came from and addressing any potential biases, said May Winfield, global director of commercial, legal and digital risks at Buro Happold. “I think we have a very bad habit of blindly trusting technology,” she said. Instead, kick the tires and do your homework to understand your incoming data and how it was produced.

Panelists remained optimistic, though, about what properly managed AI can do. "There's hardly any area of construction that wouldn't benefit from some AI in the mix, whether it's about analyzing the data, organizing it or creating new data in terms of designs,” Tortilla said. “There’s an incredible diversity of the types of data that modern AI models can handle, especially when we talk about these foundation models that have popped up in the last few years,” she added.
 
"For an industry like ours, which has this wild diversity and complexity of data, I would argue that this is the first time we've had a realistic path to bringing that together, using some of these models and tools," she said.