A hot topic right now is the use of AI in design and art, and we here at BWC have been using tools like MidJourney and Stable Diffusion for generating inspirational images when exploring an idea or design.

There’s lots of talk about what AI and Automation will and won’t replace within the design world, and from having given talks on the subject in the past and being heavily involved with automation for furniture production, I thought we’d share our current thinking on the subject.

Looking through the lens of the ‘seven steps’ we use at BWC to describe and track the whole end-to-end process of creating a building or space, here’s how we think AI will impact each step.

Inspiration

What some call ‘schematic design’ or ‘concept design’ we call Inspiration. It’s the first step where we’re working closely with a client to determine what we’re going to build to best fit their needs, the preliminary costs of the project, and what the project could look like. We tend to create several different design options, 3D models, interactive walkthroughs, mood boards, and more during this step.

We think this step is both going to be the most impacted and the least impacted by AI. 

It’s already making a big impact on generating visual ideas of ‘what the project could look like’. There are currently tools where you can submit a 3D model or photo and have an AI image generator do different versions with different materials, styles, and looks. Soon I think you’ll be able to submit photos of a building, space, or empty site, prompt for different styles, and get back lots of different actually workable ideas complete with plans and specific product lists. You probably won’t need a designer anymore for this part of this step, for anyone will be able to prompt a system to ‘show me what my house would look like with a modern addition that contains a bedroom and bathroom’. And with tools like Testfit and others, we can even see that the ‘schematic design’ of more complex projects could possibly be fully automated as well.

However, on the opposite end, a huge part of the Inspiration step is the insight, vision, and excitement a great designer brings to the table. Being able to really understand what the client wants and making sure it’s being addressed in the best way possible is still a very ‘soft’ skill. For example, we have lots of clients tell us that they love ‘Victorian Homes’ when in actuality what they love are big windows, tall ceilings, colorful exterior paint jobs, and cozy rooms dedicated to a specific use. They actually don’t want a real Victorian, which would be gray on the outside, a psychedelic riot wonderland of overlapping colorful wallpaper and trim and carpets and curtains and rugs on the inside, and a whole lot of money spent on historic reproductions and overall upkeep. Or a client wants to create an inspiring and stunning space for a group of people to gather, which requires really understanding what those people would find inspiring and stunning. Truly understanding what a client is looking for, how people respond to things, and helping get to the best idea possible isn’t something you can do with ‘prompt engineering’ alone. You can’t get what you want if you don’t know the right questions to ask to get it!

Configuration

This step is sometimes called ‘Design Development’ or ‘Concept Development’, it’s where the concepts from the first step are taken and broken down into designs detailed enough for accurate quoting and/or actual products that will be ordered and used.

It’s going from the idea of a new dining table for the dining room as part of the project to specifying a specific one from a specific brand in a certain wood and size.

We feel that this step could probably be automated at some point, where a trained model looks at your concepts and then generates plans and lists of the actual materials, products, and construction methods you might use to make the project; but that it’s more likely new combo ‘design + market’ tools like Canoa will move into this space with ever-better and smarter tools for this step. You’d be using an AI model, but via a ‘smart’ design tool that’s helping you make good selections by watching what you’re doing and predicting what else would be a good fit. Just as most good construction estimators use a combo of historic data and hard-won personal wisdom / experience to generate accurate price predictions, I think AI is likely to only empower most good professionals for this step with better data and tools to apply their experience to. There’s lots of problems in computer science where even the very best systems are only slightly better than a normal person, and not better than an expert, when it comes to certain complex multi-variable and abstract tasks. So you might be able to make a AI model that can do reasonable configurations, just like you can make one that can make a basic logo for a business, but you probably won’t be able to make one that’s significantly better at it than a trained, experienced professional who’s using AI-empowered tools.

Rationalization

This step is sometimes referred to as ‘construction drawings’, ‘shop drawings’, and/or ‘DfMA’, it’s where the configured designs are fully detailed out to where they can actually be built.

We feel that while this step is going to be greatly aided by automated AI systems, much like with Configuration there’s still a large domain of problems that aren’t deterministic enough or have too many possibilities that pure AI systems just might not be much better than a skilled human using AI-empowered tools. Engineering is still a balance of art and science, of figuring out the best compromises and solutions. So while I think automation and AI will have a huge impact on this step, it’s more likely to be in making the tools the designers are using to rationalize the work into properly manufacturable designs more productive for them to use, and not replacing the designers altogether like in some other steps.

Fabrication & Installation

These steps are where the actual construction of the project happens. Parts and materials are made and/or ordered (fabrication), then delivered to the site, modified as needed, and then assembled into place (installation).

We think this step could only be fully automated if the actual construction was so standardized and/or deterministic as to make the actual assembly easily done by robots. And from personally working with robots, nothing with them is actually very easy. Much like computers in general, they are amazing at tirelessly doing exactly what you tell them to do, over and over, but you quickly get into diminishing returns when you try to fully automate something completely.

What we think is more likely to happen is that the tools themselves used to create the parts for the project will get smarter and better, and the tools to help coordinate their assembly will get better, but much of the physical work will still be done by people. Semi-automated tools working together with skilled people will make the projects easier and faster, automated tools won’t replace the people for this step. A great example of this is all the awesome work going on in automated layout, where a robot draws or marks out where the walls, ducts, electrical, plumbing, etc. goes on the site, directed from the BIM / CAD models. The affordable and small robot makes the people doing the actual construction work more effective and productive, less likely to make mistakes, and more likely to see problems before they get too far into the project. Instead of trying to ‘solve’ the entire problem of construction assembly via robotics, even really smart AI-empowered robotics, which would cost a fortune and still not be 100% right, I see much more success in smaller, cheaper robots / AI that is just looking to empower the skilled workers already present.

Utilization

This final step is where the project is complete, and is now being used, and hopefully being studied for how well it’s actually working and/or minor follow up edits done to improve how it’s working.

This step is one where AI-driven automation could make a huge impact as well in helping identify problems and assess the performance of the project in ways that are currently too expensive to easily measure. Just a few simple cameras, cheap sensors, and a smart trained model could do what you’d normally have to hire expensive consultants to manually do. Or could make the tools that larger retail and entertainment companies already use in this space more accessible to smaller businesses that might want to measure things like ‘engagement’, ‘customer satisfaction’, a building’s energy footprint, or other post-occupancy studies.

The Robots are Here…

So the genie is out of the bottle, and AI-driven tools are here to change the construction industry in some big, bad, and awesome ways over the next decade, and I think it’s only going to accelerate from here.

Personally I’m very excited about it all! But don’t hide in ignorance and think your particular role within the construction industry is ‘safe’, instead I’d say to work on those ‘soft skills’ in understanding people better along with learning how to leverage these amazing AI-empowered tools to meet those people’s needs even better than before!