Digital Builder Ep 77: What is AI’s Proper Place in Construction?

What’s your take on artificial intelligence? 

Ask this question to different people, and you’ll likely encounter a range of reactions—from enthusiasm and excitement to skepticism or even fear. And it’s a bit ironic that a topic as technical as AI could be so emotionally charged.  

But whether we like it or not, AI is here to stay. The best way to harness its potential is to approach technology with a clear, neutral, and open mind. 

Which is precisely what we’re doing today.  

In our latest episode of Digital Builder, we’re joined by Tonya Custis, the Director of AI Research at Autodesk. Tonya and her team are deep into the world of artificial intelligence, focusing on how it can enhance how Autodesk supports its customers. If you’re curious about how AI is shaping our industry, you won’t want to miss this episode. 

Watch the episode now 

What is AI’s Proper Place in Construction?

You can also listen to this episode on Apple Podcasts, Spotify, YouTube, and anywhere else you get your podcasts. 

On this episode 

We discuss: 

  • The ins and outs of AI, along with its subsets and specializations—including machine learning, large language models, and generative AI. 
  • How we can overcome the fears surrounding artificial intelligence.  
  • The three levels of AI success.  

A level set on AI, machine learning, large language models, and generative AI 

AI is a broad (and rapidly growing) field with subsets and specializations. So, it only makes sense to start the conversation by clearly defining and differentiating AI from its other components—specifically, machine learning and large language models.  

“Generally, we will define artificial intelligence as capabilities being done by a computer that are typically thought of as being done by a human,” explains Tonya. “Tasks that a human usually does are done by a computer. It’s a wide definition.” 

On the other hand, Tonya describes machine learning as a subset of AI where algorithms are designed to learn and make decisions based on data. 

“Machine learning is when the computer learns from data. It’s learning patterns, probabilities, and other information from the data itself. Categorizing information and recommendation algorithms are examples of machine learning we see every day.” 

Meanwhile, “Large language models are models built with machine learning, and they are built on huge amounts of data.” 

And that part about data is where things get exciting because we are now at a point where we can do things at scale.  

As Tonya puts it, “The training objective of large language models is to predict the next word, which sounds so boring. However, when you can do that—and do it at scale—you also get a ton of other tasks for free. You get things like question answering, summarization, even answering math problems.” 

“That is something that surprised all of us because we had never trained at that scale before.” 

What about generative AI? 

Tonya also brings up generative AI, which represents the next level of artificial intelligence. 

She explains that AI is traditionally about encoding things. Generative AI decodes those encodings to create new content or solutions. 

“Basically, generative AI is when you put in a prompt—like a word, drawing, or image—and then the output of the algorithm generates the output for you. It’s not a search feature where it’s just finding the output; it’s actually generating. We see it a lot with language models like ChatGPT.  

She adds, “We will see it more and more because generating things saves people time.” 

One use case of generative AI in construction could be taking pictures of a jobsite and generating possible code violations. 

Should we fear AI? 

Depending on who you ask, AI can conjure up images of robots plotting to take over the world. This isn’t all that surprising, considering how Hollywood has depicted artificial intelligence (think: films like Terminator or Ex Machina). 

However, Tonya says this perception couldn’t be further from the truth.  

“Honestly, computers really are quite stupid, or I would not have a job. Number one, they don’t learn what you don’t teach them. A lot of the Skynet and Terminator scenarios are from movies, and it’s also a little bit of psychological projection. Whenever someone is afraid of AI doing bad stuff, I’m like, ‘Well, what bad stuff are you going to do?'”. 

She continues, “Technology is neutral. It’s the people behind it that do bad stuff. You could use databases for a lot of bad things, but again, it’s not the database’s fault. But we tend to anthropomorphize AI and treat it like a person. There is a weird psychological aspect to it, but candidly, AI is not scary.” 

How to demystify AI 

Misconceptions can be persistent, but there are steps we can take to demystify and clarify the reality of AI. 

For Tonya, it all starts with education. It’s about “explaining to people that it’s just math; it’s not scary. It’s explaining to people that data gives a company the competitive advantage in AI.” 

In addition to education, Tonya says we need to be more transparent about using artificial intelligence. 

“Transparency is “ensuring people feel that their privacy is protected. Or, when they’re interacting with the tool, we should be transparent that, ‘Hey, this is AI’ or ‘This recommendation was generated by a computer.'” 

The 3 As of a winning AI strategy 

Now that we’ve nailed down the fundamentals and myths around AI, let’s move on to something more actionable: how exactly can you win with artificial intelligence? 

According to Tonya, AI success happens when you go through three levels: analysis, automation, and augmentation.  


Data analysis is foundational because it lays the groundwork for all AI applications. 

“You can’t do AI without data,” Tonya remarks. 

“Capturing how the data moves through the workflow is such an important part of AI. Again, you’re predicting things. You want to be able to predict what’s next, and you can’t predict what’s next if you don’t know what will happen in the design workflow.” 

Data also provides valuable context for AI software, and the more info and context you provide to the computer, the better it can understand and predict. 


Already have your data ducks in a row? The next step is automation, which involves streamlining repetitive tasks and freeing up teams to focus on more creative and fulfilling work (Or, at the very least, getting the job done sooner and coming home early). 

“Beyond analysis, we can automate tasks. These can include tedious jobs that designers don’t want to do, or that take them a lot of time.” 


Finally, there’s augmentation, which is the ultimate goal.  

“We want to make the computer better, but we also want to make the human better,” says Tonya.  

Firms achieve augmentation when they organize their data, integrate AI into their workflows, and find ways to use the technology as a collaborative tool to complement—not replace—people.  

New podcast episode every week 

Digital Builder is hosted by me, Eric Thomas. Remember, new episodes of Digital Builder go live every week. Listen to the Digital Builder Podcast on: 

or wherever you listen to podcasts.

Eric Thomas

Eric is a Sr. Multimedia Content Marketing Manager at Autodesk and hosts the Digital Builder podcast. He has worked in the construction industry for over a decade at top ENR General Contractors and AEC technology companies. Eric has worked for Autodesk for nearly 5 years and joined the company via the PlanGrid acquisition. He has held numerous marketing roles at Autodesk including managing global industry research projects and other content marketing programs. Today Eric focuses on multimedia programs with an emphasis on video.