Skip to main content


How Far Can Automation Go?

May 09, 2018

Automation has largely been an efficiency tool, making repetitive and mechanical tasks go faster and more accurately. The introduction of artificial intelligence, virtual reality, and machine learning is starting to change the equation. Will architects and engineers eventually be “automated” out of a job—or perhaps into a new kind of job?

For the 2017/2018 Thornton Tomasetti annual report, we convened a group of specialists to discuss the challenges, limits and opportunities of automation in AEC. Here is an unabridged version of their discussion.

Phil Bernstein, Lecturer, Yale University

Jesse Devitte, Managing Director and Co-Founder, Borealis Ventures

Josh Emig, Head of Research & Development, WeWork

Stacy Scopano, Vice President of Innovation, SKANSKA USA

Martha Tsigkari, Partner, Applied Research + Development, Foster + Partners

Stephen Van Dyck, Partner, LMN Architects

Moderator, Rob Otani, Principal, Thornton Tomasetti

ROB  Automation has been part of the AEC industry since CAD appeared more than 40 years ago. With AI, VR, the internet of things, machine learning and robotics, are we now at a tipping point in its influence?

PHIL  Yes, but not because of AI, VR, the internet of things, or even machine learning. It’s more due to so many things and processes being digitized in AEC.

STEPHEN  There are two different issues. One is the automation of work, and the other is intelligence or the use of cognitive-like computational behaviors. We’ve been automating stuff for a long time — the printing of books since the Gutenberg press and the drawing process since CAD came around. And BIM is an automated tool. There was a tipping point for automation to some degree in the Great Recession, which forced a lot of people into the BIM world. The tipping point for cognitive-like computing is probably going to come when another major economic disruption forces people to get there.

JESSE  We’ve all been on the journey from CAD to BIM, and as investors, the question is: What is the next phase? We think it’s connected data. We don’t think of this as a tipping point so much as a unique moment in time. We spent the last year talking to a lot of people in architecture, engineering and construction firms about the future of investment in this industry. Many of them are asking themselves the same question. All of these connected data technologies are arriving on the scene simultaneously. Everyone’s trying to figure out whether they’re applicable and what they mean for their business, their practice, and their customers.

STACY  From the construction side, this is a time of churn, not of tip. We have new ingredients on the table, an explosion of startups and the death of startups. When the tipping point is really coming, you will see true digestion of those ingredients and their convergence. Right now I just see VR for VR's sake. I see 1,000 opportunities for AI. And we’re really unpacking what people mean , on a spectrum from machine-learning applications to true artificial intelligence. Almost every technology I've dealt with has bounced over to the customer space and they’re asking, “Here are a bunch of hammers. Can you tell us where the nails are?” This is why I get a sense of churn.

ROB  Speaking of an economic forcing, everyone's margins are not what they used to be, even when times are good. We are as busy as we have ever been, but margins are lousy. It's as if the building industry doesn't operate in the framework of Keynesian economics.

STACY  The economics question is a quickening for a true tipping point. There will also be a quickening of the technology vendors. Ventures and startups will be overextended and just won’t be able to survive when the dust settles. What comes out on the other side will be those resilient concepts or organizations that can really support this synthesis.

PHIL  The biorhythms of the tech industry and of the AEC industry are completely asynchronous. The tech industry works on a cycle of three to six years from investment to go-to-market. The cycle in the building industry is longer– projects are three to 12 years, and systemic change takes 8 to 15 years. So the churn happens where these two sine waves overlay and interfere with one another.

MARTHA  The rhythms are different because technology is usually not seen as an intrinsic part of the industry. Some companies, like Foster + Partners and WeWork, have identified this gap, and have teams that include programmers. You need to have the technology to support that time scale, and you need to make sure that you keep up.

ROB  How do we get people excited about this technology and address the pitfalls that hinder adoption of automated methodologies?

JOSH  One pitfall is overcoming the perception of threat. A recent McKinsey report* said automation is going to create more jobs than it destroys in the construction and infrastructure sectors.

JESSE  From an investor’s standpoint, going mainstream and finally getting McKinsey’s attention, is a good thing. We used to be talked about only in ENR. Now it’s also Fast Company and Wired.

ROB  We went from CAD to BIM and design of fabrication. There are still people who believe in a manual way of doing things. In a way they’re right, because what we used to do 20 years ago on 25 sheets now takes 200 sheets.

JESSE  As investors, the other thing we look at is how your customers are changing. WeWork is exhibit No. 1. Talk about a major disruption to the real-estate owners who fund and build and operate these buildings. They’ve seen their business model upended by the likes of WeWork, and that’s a change in how we live and work.

PHIL  We need to remember, however, that those of us around the table don’t really represent the AEC industry. In the U.S., it’s mostly firms working in small towns and employing five-to-seven-people that make up the 750,000 companies that do construction in this country. The reason our industry moves so slowly is because it’s simultaneously huge and tiny.

MARTHA  Yet – and Foster + Partners aside – it’s the small firms that tend to be more innovative than the big firms. Change is much easier for a small team of people who can just go ahead and do it, rather than a big company with a lot of investment in legacy systems and processes.

JESSE  And that is more technology reliant in many ways.

PHIL  All I’m saying is the demographics are extremely varied. For every person employed by a cutting-edge firm, there’s somebody in Poughkeepsie using a copy of AutoCAD 11.

MARTHA  You’re absolutely right, but every industry is like that. And yet, other industries manage to move forward with the times much faster than the construction industry. Firms that recognize that new technologies enable them to be more creative have managed to outpace firms that resist change. And the difference may not be generational. I sometimes see Baby Boomers who are more forward-thinking than younger people.

JOSH  Part of our industry’s issue in the United States has to do with our over-privileging of higher education and our underprivileging of trades as a legitimate path in life. As a Swedish company, I’m wondering if SKANSKA has a different perspective, if you invest in trades and move people in that direction.

STACY  In Scandinavia, the trades are more organized and the unions more dominant, whereas in the U.S. they are shrinking in number. Across all the trades, one of the primary remits is the quality of the workers. Today we have a bunch of specialists organized by discipline and specialization for each part of the project life cycle. What we see now is a shift from seeking the I-shaped (deeply specialized) individual to the T-shaped individual (specialization with broader digital acumen). In SKANSKA, real innovation happens when we force-fit two I-shapes–a young er, technologically capable person with someone who has years of practice experience. We have a young lady who came to SKANSKA from architecture, migrated into Precon, and we sat her next to one of our senior estimators. Now they are batting near 1,000 in their pursuits. She can immerse a client in design options, powered by every breakdown structure and every cost. Every “what if” scenario is automatically precalculated. These two together can offer exactly the service that we should be providing to our clients, and over time you won’t need two people to be doing that. You’ll end up getting both in one. You’ll see the tipping point when you see that type of personnel shift.

PHIL  The problem you’ve identified, Josh, presents two essential challenges to the industry. The first is that there isn’t somebody saying, “We really need to upscale all these ironworkers and sheetrockers.” It’s left up to individual firms, but the profit margins are so thin that nobody can invest in that sort of training. The second problem relates to network theory: it’s well understood that when you have a loose network like the building industry, where the dependencies are episodic, they’re not strong. When you introduce innovation in one node of that network you actually degrade the efficiency of the rest of the network. These are structural reasons why innovation doesn’t diffuse well through the network. And these are people issues, not technology issues.

ROB  Who drives change? The industry, the customers or both?

JOSH  There was a phase, especially around IPD, when we talked a lot about finding enlightened clients and enlightening clients. We considered it our job to help the clients understand this process. I think that’s fundamentally wrong. There are clients who love the process, but they’re the exception. Everyone else wants a hole and not a drill, and we should just deliver something that’s consistent and predictable.

STACY  In the “hole not a drill,” scenario, we’re usually in the way. If you ask a client what’s getting them out of bed in the morning, it’s their operations, their core business, and we’re providing a facility for that business. And until that facility is complete we’re preventing them from doing what they’re organized to do, which is some other business–provide health care, provide office space.

JESSE  In my view, the owner-operators are driving change because they can understand the longer term impacts. Look how far health-care facilities and clients have come. Intel always cared about its plants, as you would imagine, so they vertically integrated to drive that process. Now health-care systems have gone the same way. Where you see a drive for vertical integration, you don’t have to scratch too deep to find innovation.

STACY  Our experience in health care is the same. We see the IPD-type relationship baked into the partnership model early. We have the license to look for value creation from the onset, instead of value engineering way late. That creates a sandbox where I think all the tech starts to fall into place to deliver what Josh said—something consistent and predictable.

ROB  How will AI change automation—and beyond? How far can we go with this?

PHIL  The father-son team of Daniel and David Susskind, from Oxford University, have written a book, The Future of the Professions, in which they assert that machine learning and artificial intelligence will eliminate the need for professionals within a generation. You won’t need architects, engineers, lawyers. It’s already started. An automated radiology algorithm can now read a regular c-ray better than a doctor can. Sixty million disputes on eBay were resolved by an algorithm. A lot of work that architects and engineers do is straight production: that’s going to go away first. Then basic decision-making is going to go away—you structural engineers are probably the first to go because you’ve been so rigorous about proceduralizing all your knowledge and your work.

STEPHEN  When we talk about AI, productivity and the effects of cognitive computing, we need to separate A and E and C, because each will be affected differently and at a different pace.

JESSE  There is no shortage of AI start-up ventures working to get funding around the engineering space, because precision and procedure are highly valued and amenable to cognitive computing.

STACY  Exactly. Anyone, like engineers, who has good structuring of their information, from accounting to codes, will most likely see the earliest signs of automation. Those professions and roles that generate or consume unstructured information will be more challenging to automate early.

PHIL  I think that’s a distinction without a difference. We’ll eventually get to the point where the existential question becomes, How much of the building industry across all three silos will society accept as producing a routinized thing? Because when I can look at 900 hospital models and extract 85 percent of the best decisions, I need an architect to do something else.

STACY  The point is, in what phase and what order? Structured information is ripe for automation powered by AI. Nuanced market conversations where mining tweets and general consumer or industry trends adds levels of complexity, those will probably be the second or third phase.

PHIL  The existential question is on the AE side, not so much the C side. Legally, there has to be a human structural engineer and a human architect who sign and seal a drawing or certify that there’s professional competence, and the reason for that certification is the public’s health and safety. Once the public’s health and safety has been automated, and society works out that problem of algorithms taking the responsibility for things like surgery and reading x-rays, then architects and engineers are not at the table anymore since owners won’t need us to sign and seal drawings.

MARTHA  I don’t care if AI takes over everything. I will leave architecture and go live on Mars [laughter]. As with the Industrial Revolution, the AI revolution is here. We accept it. I don’t think it is a bad thing. I think it is an opportunity to do other things above and beyond what we’ve been doing. We have been using AI in our office for a lot of interesting things. We have also been using genetic algorithms (which are not AI) for many years. We’re now using neural networks to solve a plethora of interesting problems that negotiate huge datasets – a task that a human would not be able to achieve nearly as fast. And we’re doing it to good effect. In 10 years, will a computer replace me? I hope it will, because for me personally, it’s an interesting challenge: where do we go from here?

STACY  Once automation starts to complement a better understanding of business, you get leaner. The shape of activities changes dramatically. At some point you have to go from digital to physical, and luckily that’s our side of the fence.

MARTHA  The interesting thing with neural networks is that we ourselves do not understand how they work, but they can train themselves to evolve to such an extent that they can traverse big datasets and make connections that we cannot make. I think that’s a good thing, particularly when it enables automation.

STACY  The common thread here is that AI-enabled automation shifts our day-to-day objectives. On-time and on-budget becomes a low-level given, and now we’re targeting higher client/team goals like productivity, customer or employee satisfaction and wellness objectives.

ROB  Given that A, E and C will adopt AI differently and at different paces, how long before, say, 50 percent of what we currently do is automated?

STEPHEN  I don’t know if I can put percentages or times to that. The change is going to come faster than it ever has before. We had 20 years to absorb BIM, for the draftsman to either adapt or to find something different to do. We don’t have that time with AI, and that’s why the existential conversation is so prescient. The rate at which we will have to adopt it will be faster than we can comprehend right now. It’s scary for people. That’s why we need to begin to deal with these issues.

PHIL  The book Only Humans Need Apply explores what happens sociologically when large swaths of society’s work are automated. There are only so many rocket scientist jobs available, and at some point there won’t be enough work for everyone to do because it’ll all be automated. Right now, I’ve got 300 graduate students who are going to have to be working, and 70 new ones every year. Is that sustainable?

JOSH  That was the upshot of a McKinsey report** from last year. There are things that resist automation because of a lack of predictability and repetition, like trimming the tree limbs around your house and home health care. And the market still serves as a damping mechanism to change in many places. Thornton Tomasetti might be 50 percent automated in five years, but a less technically advanced engineer might take 10 or 15.

PHIL  If I said to your management committee, “In three to five years you will not need 40 percent of your staff doing what they do now. You can repurpose 10 percent of them,” what are you going to do? You’re going to be 30 percent smaller, doing the same amount of work.

STACY  Technology shifting jobs is an old notion. The industrial and internet revolutions did eliminate some jobs. But in total, roles evolved and the nature of work advanced alongside these evolutions.

STEPHEN  I don’t think AI is going to replace our ability to negotiate with communities and various stakeholders. I don’t think AI will replace – at least in our lifetimes – the ability to triangulate all the complicated factors that go into making a successful project.

PHIL  Are you willing to do it with 40 percent fewer people? If you look at the demographics of the American architecture profession during the recession, it contracted by about 30 percent. Revenues went way down. By 2015, net revenues of American architects were back up to pre-recession levels. Yet 15% fewer people are working in American architecture now. And why is that? It’s BIM. It’s the only thing that changed.

STEPHEN  Absolutely — we’ll do it with fewer people. We are doing it with fewer people today. We’re building way more square feet every year than we ever could have 15 years ago, thanks to the technologies that you’re talking about.

JOSH  This might sound overly idealistic, but what if, instead of figuring out what you would do with 40 percent fewer people, you figured out what you could accomplish if you kept that 40 percent, above and beyond what you do now?

STEPHEN  I think that’s what technology has done, at least during my career as an architect, in the last 15 years. Every time we’re better or faster or something, it’s because the technology lets us do more. So our end result is better. As long as the adoption process is quick enough, it’s easier for us to say, “OK, we’ll just do more now, not lay people off.” But I agree those economic cycles do drive the larger systemic problems that you’re talking about. I think in the next downturn, that’s when you say, “Well, maybe we don’t have time for those 40 more studies that we’re going to do with this 40 percent more labor.”

JOSH  But we have a window then. Those truck drivers who are sitting in self-driving trucks “just in case” are doing Coursera while they’re riding in the truck. Because in 10 years, no one will be a truck driver.

MARTHA  We all find ways forward. When I joined the office 12 years ago, we were already heavily into R&D and automation and 10 people were doing it for a firm of 1200 people. Now a far bigger percentage of the office knows how to program and use parametric tools. Does that put me out of job? No. It forces me to look into what I can do better. Nowadays, even our more skeptical designers understand that computational design techniques may enhance rather than impede their creativity. Will computers take over in terms of creativity at some point? Maybe they will, and at that time we will think of something else to do. I think of this as a potential job change, not job elimination.

PHIL  What we don’t do on the A and the E side is try to design our future, because we operate at an order of magnitude smaller than the rest of the forces in the industry. One of the lessons that came out of Only Humans Need Apply was that you have to design a future state before it is created for you. Those of us who lived through the CAD revolution of the nineties might remember the short period when we were charging for CAD. Then, clients said, “Oh, you’re 50 percent more productive,” and fees got cut across the board. So you won’t have the extra 40 percent to play with. Your fees will go down by 40 percent, unless you come up with some other value proposition.

ROB  Architects and engineers are only five to 10 percent of the project cost. Construction is by far the biggest chunk. Stacy, where do you see automation going?

STACY  By far the biggest issue in our industry is safety, and it is an enormous focus at SKANSKA. Across the industry in the U.S. alone, we average three to three and a half fatalities per day. A lot of the technologies we’re focused on are for improving safety. For example, we’re renovating the old post office near Penn Station. Because of sightlines, the operator in the tower crane can’t see a large part of the project. We typically spend $1,200 per radio headset to connect the cab operator with the guys on the ground. All we did was find a head-mounted camera that can connect over Wi-Fi to give him a boots-on-the-ground view of the pick. And this was revolutionary. Now the ground guys can say, “Phil, I need you to look right just a little bit more, because you need to see what’s happening with that strap.” It’s such a low bar when we take simple capabilities and throw them at real site challenges. There are just a lot of nuggets out there like this.

STEPHEN  I would say the bigger question for you is why is a computer not running the entire thing? And clearly an answer to that, to put words in your mouth, is the unions, labor organizations and everything else that is standing in the way of this kind of progress. We have had the technology to completely automate crane picks for 15 years, maybe more. So it’s fascinating to me — that’s a great example of a great little technology, but it’s solving a tiny problem that probably should have been solved in a much better way a long time ago.

STACY  True, we control the construction site, so you’d think we could do this. In reality, the crane guy is a separate trade contractor and we have to balance how we define requirements so we enhance (not hinder) their operations.

STEPHEN  We are, to some degree, dictating means and methods now. That’s the future of designing and making buildings.

ROB  Phil brought up the idea that structured data makes engineering more amenable to AI. For the architects and contractors — how do you structure and capture that data?

MARTHA  Sometimes we capture data systematically. We directly classify it so we can use it for a neural network. But sometimes most of the data is not captured. WeWork is in a fantastic position because you’re your own client, so you have the opportunity to measure what you’ve done and assess how well it works without requiring permission for post-occupancy studies, as we do. And if you use IoT systems within your offices to capture that data directly, then it’s easy to assess what the next steps should be. What should the desk distribution be in a space if I need that much light or if I have so many people? If the data you have captured can be analyzed, then space planning becomes a straightforward process.

ROB  When you talk about AI it’s all about data. At our firm, with thousands of projects, it’s actually not a lot of data, even including the Revit model.

MARTHA  Collecting post-occupancy data is one way to do it. Another way is to generate your own data. We have done this by randomizing configurations and, using bespoke programs, creating huge datasets with different performance criteria that we’re interested in, literally within half a day. We can create hundreds of thousands of solutions that we can then use to train our neural network. For me, the most interesting proposition is when you are able to collect that data from your client, and offer a solution as a service.

STACY  We’re completely modernizing our foundation for data collection. As a 125-year-old international company, we have a lot of legacy (and duplicated) technologies — 80 HR platforms, for example. So we have to take a big step back and ask, “What data do we have and where is it?” The next major component for us is figuring out what makes us profitable or what correlates with profitability. That’s a huge question. The answer might not be just in our data. It might also be in OSHA safety data versus SKANSKA safety data, for example. You’ll need analysts and data scientists (internal or external) to make sense of this. We can build machine-learning algorithms coached off the data science. As for the AI engines, IBM’s got one. Google’s got one. Microsoft’s got one.’s got one. And they’re all being decoupled from their core businesses and asking us: “What are you going to do with this hammer? Do you have any nails?” We’re trying to figure that out. Did we define any nails that are really good for Salesforce but not Watson? Each of those has different strengths and weaknesses. That’s why I echo back to the churn conversation. It’s what we need to do to make ourselves ready.

JOSH  So many of those conversations–about machine learning or block chain — they all start the same way: “We have computer vision. What do you guys need?” And we say: “What do you mean?” That’s not usually how you bring a product to the table. They want to know what our nails are.

PHIL  They have an intuition but no factual information that this thing they’ve made may be useful. This echoes Jesse’s and Stacy’s observation about churn. There’s all this stuff out there, and it’s like Tinder (the dating app) for AEC technology.

JESSE  On the other hand, data could change the very nature of the industry. I was in a real-estate tech meeting recently and the discussion was: “Will we need less real estate in the future?” The conclusion was, “Yes, because of better data, we are continuing to learn about the inefficiency of it. We know we can do so much more with what we have, so we’ll naturally build less.”

STEPHEN  Thanks in large part to computer vision enabling a better understanding of componentry and existing conditions, new construction is going to be a different percentage of our design portfolio in the coming years. We may work with a completely different kit of parts. Maybe not new steel. We might be just saying: “We’re going to reconfigure what we’ve got here.” A machine can catalog that. It’s a completely different marketplace based on some of the tools. Triangulating the possibilities from guys who are saying, “Hey, I got this thing. It does this.” And another guy who says: “I got something else.” And clients who are saying something else. It’s now our job to define the problem space for the growing list of these new tools.

MARTHA  I think many vendors and developers are interested in getting into our space but sometimes it’s a matter of the industry being too small for their interests or the entry cost being too high. Nevertheless, we need to create opportunities to let the ideas flow because often ideas with great potential value to our industry come from outside our domain.

JOSH  People are coming to us for different things. Usually they’re coming to us about our core business, which is co-working. Our challenge today is that we now have more than 200 locations, and so anything that we want to deploy in a physical space, we have to do it in 200 locations globally. It’s not: “Let’s do this building.” It’s more like: “Let’s do this worldwide.” That requires a completely different set of criteria, and it becomes a litmus test for a lot of vendors that come to us: How much do they understand their own product, and how much do they understand the implications of working with us?

ROB  So there are all these people trying to sell to us, and many new ones trying to invest in us. Jesse, what makes the AEC industry attractive to VCs?

JESSE  It’s obviously large and the last major sector of the global economy yet to be transformed by technology. It appears to the outside world as very inefficient, fragmented and under-connected. WeWork has had a major impact and created an extraordinarily valuable company. Consider Airbnb and Uber as well. These are all connected to the broader built environment. People have begun to realize this is the hunting ground for the next version of one of those companies.

ROB  Have you seen any companies trying to capture the data of all of building stock, new buildings or existing buildings?

JESSE  Measurable, in San Diego, has north of 5 percent of U.S. commercial building energy-usage data, and their clients are real-estate investment trusts and others who are trying to understand the performance of buildings. After the downturn, real estate investors said, “We’ll never accept the black box on real estate again. We want to know what’s happening at an asset and building level including performance.”

ROB  CASE*** used to have a saying: “Buildings equals data.” Do you think there needs to be a universal way that building data from new and existing buildings can be required from a city standpoint? It would be unbelievable if engineers and architects had to distribute all this data into their files.

STEPHEN  That would be like asking every fund manager to release their portfolio as well as how they trade. It’s the secret sauce.

JOSH  You could probably mandate release of a subset. One of the Big Data stories from a few years ago was how New York City changed its approach to code enforcement and other things, because of the analysis of information that had been digitized from routinely required forms and filings. There is still a lot of dark data out there that is not captured, aggregated and analyzed. But forget about mandating submission of anything proprietary.

STEPHEN  That’s like saying, “We’re not going to give you the useful stuff.”

MARTHA  I can envision apps that will soon be able to capture that. They already do; we just don’t know it. For example, what is Citymapper doing? It records everything I do and when I do it, and creates revenue by selling that. I find the app useful, so I accept that my data is going to be used in such a way. I believe this is going to happen in the built environment too.

STACY  We’ve all opted in. I heard a statistic, which may already be dated, that in a month’s time we sign and authorize as many contracts as our grandparents did their entire lives. We’re coughing up our data everywhere. It’s becoming second nature. Google has access to every Android phone’s GPS location and every Google map user’s start-stop mode of transportation. They can pull that up in real time, and that data is unbelievably valuable. From a developer’s standpoint, where are the bottlenecks in traffic where people start and stop? Where do they want to be? These questions can be easily answered by access to this level of “private” data.

MARTHA  The profession of the future might be statistician, because it won’t be enough for you to have all that data. It’s about how you traverse it. Even if you want to build a neural network capable enough to do these things, you need to be able to statistically analyze the data you have captured.

STACY  If I were advising my college self today about an AEC career, I would say: “Coursera the hell out of data science.” Whatever career I might pick, it would have to be augmented with data science.

STEPHEN  I would argue the other way. You need to remain focused on the big-picture issues and needs of society. You can be a technician with all the computer science you might ever need, but you’re still a technician who has to adapt continuously. The constant is that we’ll still be trying to address these societal needs. To me, as an architect, that’s the value we need to provide.

JESSE  My advice to my present-day college self is urban planner. That profession is going to be even more important in the future.

JOSH  My master’s thesis in 2005 was on data. I didn’t produce a building and all the architects thought I was out of my mind. One of my professors said, “This is the most boring project I’ve ever seen.” Data science wasn’t new at the time. I took a computational intelligence class and learned about neural networks and genetic algorithms and fuzzy logic because that’s what the engineers were doing, and they called it “decision science” because it was about making decisions. If I were advising myself today, I would do that again, but I’m not sure that it served me in the way that I thought it would at the time.

STACY  Going back to the beginning of our conversation – are we at a tipping point? I don’t know what’s next for VR/AR, or what’s beyond AI. These technologies have exploded onto the scene in the last 10 years. They all had legacies before that, and now we’re trying to figure out how to make sense of them.

PHIL  PHIL The tipping point may be a tip into an age of more rationality, especially for architects. In the 1980s, I was trained as an architect to elevate my intuition and judgment as someone who could synthesize a complex, ambiguous, wicked problem, largely unencumbered by facts. And we’re moving quickly into an age where an enormous amount of information, analysis, inductive reasoning and historical data is going to be available. The building industry has to move into a rational frame and away from privileging the nonrational over everything else. You don’t want to eliminate the ineffable, but we have to get ourselves on a more rational footing. Several years ago I was talking to the guys from an architecture form that was working for Google. They started a kind of normal design process where they came in with sketches and had alternatives. And the Google guys said: “Where’s all the data that this is coming from?” And the architects said, “What?” It completely changed their whole way of looking at the world. And it’s not just Google doing this. That concept ties to the discussion of outcome-based delivery models. You can’t just say: “I’m going to an outcome-based delivery model.” You have to say: “I’ve done enough analysis and work that I know this is going to happen. I’m going to make a promise that this is going to happen, and when it happens, pay me. And if it doesn’t happen, punish me.

MARTHA  A lot of the innovation we’ve been pushing is about creating applications, tools and techniques to help designers make decisions in real time based on analyses that they are also getting in real time. We’ve been doing performance-driven design for at least a decade. Now we’re heading to something bigger than that: getting answers even before we ask the questions.

ROB  Looking over the changes we’ve discussed, what steps should we be taking now to prepare for increased computation/automation?

STACY  Cultivate the T-shaped people and reverse mentor them. Pair the high-powered tech talent with the high-powered seasoned professional. Don’t try to script it.

STEPHEN  You don’t need to change the financial structure of your office. If you’re hiring right, you don’t even need to change how you’re hiring. What you need to do is empower young people. That’s a hard thing for this industry to do, because our profession has been built upon a grand bargain, that we operate in a professional society, and have to clear these hurdles. We do the internship and working punishing hours. We go to Yale. We pass the exams. We become members of this exclusive association that gives us the right to do stuff. We need to let go of the idea that traditional educational models and traditional apprenticeships are required to gain the expertise to do the work we do. Some of our best designers are young people because they have the ability to churn out 500 more ideas than the guy who was doing their job 15 years ago. They’re able to analyze those ideas, synthesize them, test them, fabricate them, and learn at a rate that is much more productive, and gives them expertise at the age of 35 that in the past you might have attained at age 65.

PHIL  There was an incident six or seven years ago where some guy, playing college football, did a move on the field that he could have only learned by playing Madden NFL, because if you’re a football player in this country, and you play all the way into the pros, you will not play more than 40 or 50 actual football games, maybe 60, in your life. But if you’re playing hundreds and hundreds of games in Madden, your brain is processing way more information. This guy ran some pass pattern and the old people reacted: “What the hell was that?” And the young people said: “Oh, yeah, yeah. You can do that in Madden.”

STEPHEN  We have to hope that people like Phil are educating our future staff to think this way, to be the young thinkers and designers that we can trust and to whom we say: “I get it — you’re only 30, but you’re much more powerful than a 30-year-old used to be.”

JESSE  Some of the firms that cultivate these young people are leading in innovation by bringing technology talent into the firm that’s not just from the profession. They invest in multidisciplinary teams that might include an aerospace engineer, an architect, a manufacturing engineer and so on.

MARTHA  That’s the heart of where we should be. Everybody likes little labels — “I have these degrees.” We all need to be like Schrödinger’s cat — not alive, not dead, not architects, not engineers, not computer scientists. We need to be a bit of everything. And to do that successfully, we need a bit of humility.

PHIL  We also need to be clear about what we’re good at and what we want to be good at in the future. A symptom of churn is the reaction: “I’ve got to be using fill-in-the-blank technology.” A better reaction might be to focus on what you’re good at now and what you want to be good at in five years, then align the way you think, hire, structure your organization and invest to pursue that goal.

JOSH  What you’re describing takes a tremendous amount of discipline. Also, you have to say: “I think this is what I’m going to be able to do in five years, but I also accept that in six months I might be doing something different.”

ROB  To deal with this change, do we in AEC have enough entrepreneurs, or are we mostly craftsmen?

STACY  If you don’t enable entrepreneurialism, you’re not going to have the workforce to reduce by 40 percent. There’s something about the generations that are coming, the millennials and the Gen Zs behind them – they are decidedly more entrepreneurial.

PHIL  Where I’m teaching, there is a surge in the number of people who want to do joint degrees: MBA and M.Arch. They want to be in the entrepreneurial environment of the business school, and they’re not getting enough of that in architecture school.

MARTHA  The entrepreneurship model that works in the technology industry does not work at all in the AEC industry because so much is based on legacy. In an entrepreneurship, you trust somebody who’s very young and give them a lot of money to develop their startup. I can’t think of an architecture firm that would give a 25-year-old a lot of money to go build an architectural practice.

JOSH  Attitudes about entrepreneurship can be taught. One of the most successful projects I’ve given students was in a joint architecture/construction program. The project was to design an AEC business. These kids started to think about what problems needed to be solved and how to generate a value proposition and how to identify market opportunities. It made them feel empowered: “Oh, I actually have something to say about this industry and about how people do this business.” Generalizing that, I think it’s important to empower everybody. Just as a 35-year-old can learn so much on Google, so can a 65-year-old. How we should prepare for increased computation and automation is therefore a people-answer, not a technology-answer. We need to spend less time hand-wringing about how technology changes our current process. Instead, we need to empower all these people to learn and take advantage of the technology, as active producers, not just passive recipients.

* "Jobs Lost, Jobs Gained: Workforce Transitions in a Time of Automation," December 2017, McKinsey Global Institute

** "A Future That Works: Automation, Employment, and Productivity," January 2017, McKinsey Global Institute

*** A consulting firm co-founded by Dave Fano, which joined WeWork in 2015.