By Roland Murphy for AZBEX
When a product, service or technology is as cross-functional and pervasive as artificial intelligence, multiple mindsets emerge nearly simultaneously in the marketplace.
Developers and investors tend to over-hype current capabilities and espouse a packaged set of sunshine, lollipops and rainbows about its future potential. Early adopters try to shoehorn it into every aspect of their operations, whether it fits, doesn’t fit or may one day fit. Detractors and luddites fear it portends the end of the world as we know it.
AI has experienced all of these events, and others, to an extent rarely seen since the launch of the World Wide Web.
We won’t waste space or readers’ time by pointing out all the various ways AI has impacted daily life in general or the planning, development and construction industry in particular. If you’re reading this, you are probably in the industry and are at least as well aware as we are.
What we plan to do here is examine one recent use set and get a feel for how the technology may help developers in the planning and zoning process, while highlighting those areas that simply cannot be farmed out to machines and tech.
Old Problems, New Tools
Phoenix Business Journal recently ran a feature on a new AI product called ChatAEC, produced by AIAEC. AIAEC was founded and is led by Ali Fakih, who is also CEO of Sustainability Engineering Group.
Simply put, ChatAEC is a development tool to help development professionals understand and interpret the various land use, zoning and development regulations covering a particular site. According to the article, “ChatAEC summarizes zoning designations, overlays and site specific constraints by entering an address or parcel number. In addition to providing zoning insights, ChatAEC also generates detailed due diligence reports for projects.”
The product is currently available in Arizona, and a nationwide version is planned for this summer. The ultimate vision is a platform that can support users across the planning, permitting, design and quality control process with tools that let users analyze sites against their design and development components.
While it can take a human researcher hours or longer to go through the applicable due diligence regulatory components, ChatAEC looks to handle the task in seconds.
Technology Limits and the Need for Human Expertise
So, where does that leave the land planners, project representatives and researchers currently handling the work? According to one Phoenix attorney, it leaves them in a pretty good spot.
We reached out to Adam Baugh, a partner at the Arizona land use and real estate law firm Withey Morris Baugh, for his insights. WMB is one of the state’s leading firms in the field and is part of the top tier of “heavy hitters” developers rely on to navigate particularly large, complex or cumbersome project approval processes.
Baugh said an AI tool can be valuable as a starting point, particularly for smaller, more entrepreneurial developers. “When you’re less experienced and more cost conscious, an AI tool is a valuable tool to start. The problem is when the AI tool, which should be a start, becomes the end-all, be-all. I think what happens is there’s a little bit of over-reliance and confidence in the tool. The challenge with that is AI should give you a place to begin and maybe help you see things you might overlook.”
He continued, “I think for that reason, it makes a lot of sense. There’s probably some cost reduction there. There are probably some areas you might be overlooking that could call your attention. It could create some speed and efficiency.”
He added, however, pulling from published public data can present challenges. Public maps and documents sometimes contain errors a trained human would recognize, but an AI screening would not. Baugh added that some tools lack the ability to find hyper-local and real-time data and cannot, for example, account for the lag between when a zoning is approved and when it gets entered on a map or when an assessor updates a property transaction. How rapidly those updates are made varies between cities and staff members.
“What it does is regurgitate out and synthesize what it finds,” he said, “but if it’s not finding the most relevant information, the most updated information, or if it doesn’t know the neighborhood-specific things that aren’t necessarily able to be pulled off an online website or staff report, it could miss some pretty big landmines.”
Another function AI cannot serve, and likely will never serve, is negotiation and advocacy. As former Scottsdale City Councilmember Tammy Caputi frequently said about rezoning and project approval requests, a project can check off every box under the requirements and still get rejected because officials bring their own subjective preferences and prejudices to considering applications.
When we ran that comment past Baugh, he added another wrinkle, in which properties could be excluded from potential development because the AI tool may look at only the expressly allowed zoning and not consider potential alternate uses that could be or have been allowed previously.
“I’m not saying AI can’t get there,” he said. “What we’ve seen over time is how quickly these tools improve and get better. I think those tools, in the sense they can give direction, are fantastic, but the output shouldn’t be deemed definitive, rather than directional. If you rely on those things, a poor assumption might result in a bad acquisition decision that’s more reflected in the pro forma that might take years to catch up.”
He went on, “I think they can be useful in the due diligence site planning context, but in the sense that they give direction and ideas, but not in the sense that they can replace the human touch.”
Baugh said he makes use of AI in his own business, but he is comfortable with the technology because he understands its functional limitations and where direct human engagement and effort are needed.
He also cautioned against AI’s tendency to occasionally generate incorrect or apparently wholly manufactured content, including court cases in which the technology provided completely invented case citations.
“It’s a valuable and useful tool, particularly for small groups, but not to the extent that it replaces decision making,” Baugh said. “It might get it right most times, but you can’t afford the time it gets it wrong.”
A Potential Boon for Plan Review
Given that plan review times are a persistent pain point for developers, we asked Baugh for his opinion on AI as a tool to expedite the review process for code compliance.
“That is a great question,” he responded. “I’m not sure that tool exists, but I don’t think it will take long for that to be out there.”
He added that maybe an ideal use for AI. “Code compliance is really clear,” he said. “It’s black and it’s white. If you want to accelerate review times with staff, put it in there. But here’s the caveat: Cities have to adopt codes and ordinances because there are thousands of properties, and you can’t write a single code for every single property. You have to write broad-based codes.
“For that same reason,” he added, “not every site can comply perfectly with those codes, which is why you need a human touch to evaluate if there is a benefit, a waiver or a technical appeal because of the unique circumstance of the site.”
Baugh said the reason the process needs a planner and a dialogue between staff and developers is to identify and work through applying practical reality to the black and white code. He said the ideal iteration, application and intersection would be where AI identifies a conflict point and flags it quickly so staff and project representatives can begin working to address the issue early in the process.
When all is said and done, Baugh’s views reflect those of commentators who have generally tried to inject some balance and rationality into both the hype and the fear.
“It’s a great tool, but it is just one factor of my overall research,” he said. It cannot pick up on issues like city councilmember biases or similar cases that have been successful or that have failed based on non-documented factors. “There are plenty of ways to use it, but there is not a singular way it eliminates what we do.”
He also said the industry has to be acutely aware of potential consequences if AI gets an analysis wrong. “In no scenario should we overlook its utility, but a lot of AI is over-hyped.”
Baugh added that his own use of AI has occasionally led to instances where the tool generated projections and conclusions that did not match his own research and understanding of laws and processes when examining an issue, which caused him to at least consider the different result presented.
“I don’t know if I’m right or it’s right, but I know that 20 years of experience informs me on how I should moderate or temper what AI is telling me the result would be.”