Augmenta: Automated Building Design Using Generative AI

Augmenta is developing a new AEC technology solution for automating building design, and what makes it different from existing attempts to automate building design — such as the tools being developed by Bryden Wood as described in the article, AEC Technology Development at Bryden Wood, or the smarts in Autodesk InfraWorks, as described in the article, Why Isn't There a Smarter BIM Tool for Building Design, Yet? — is that it is using the breakthrough technology of generative AI (artificial intelligence).

Best exemplified by the tool, ChatGPT, which seems to have taken the world by storm, generative AI is a type of AI that can be used to create new content. In addition to ChatGPT, where the AI creates text, another well-known example of generative AI is DALL-E, where the AI creates art. These might not seem like professional use cases, but technology research firms like Gartner predict that generative AI will soon become sophisticated enough to augment and accelerate design in many industries (see Generative AI Use Cases for Industries and Enterprises).

Augmenting and accelerating design is what Augmenta is aiming to do for the AEC industry.


While the application of AI technology in AEC is not new, it has mostly been used at the construction stage so far (as described in the article, AI in AEC Updates, 2022). In contrast, Augmenta is focused on the building design phase, specifically on the different engineering disciplines — electrical, plumbing, mechanical, and structural — for which it can automatically create detailed, code-compliant, and constructible designs that satisfy a customer’s specified criteria.

What makes Augmenta especially compelling is that although it is hosted in the cloud, it integrates directly into Revit with a plugin, so that designers already using Revit for detailed design and BIM modeling can continue to work within it and still use Augmenta as a smart design tool within their existing workflows. They don’t need to leave Revit or import/export any files. However, since Augmenta is a cloud application, all the logic and processing required for design generation is done using cloud resources. There is no local installation that needs to be updated each time a new enhancement is added to the application.

While development work on the different modules of Augmenta is still in progress, the electrical module has been completed and is now available for early testing and pilots, serving to demonstrate how the application would work for automating designs in different disciplines. For the electrical discipline, Augmenta Electrical automates the design of electrical raceway systems. This is something that is currently done manually by the electrical engineer and contractor and is a difficult and tedious process, requiring rerouting every time there is a design change.

Using Augmenta’s Revit plug-in, you specify the locations of the electrical equipment, other requirements and settings, and indicate the walls and ceilings through which the electrical raceway can be routed. The application then provides you with multiple raceway design alternatives that meet these specifications. You can evaluate the different options in parallel, comparing tradeoffs between cost (labor and materials), amount of material, construction time, and future ease of maintenance (Figure 2). Each design is ready to procure and build, so you can select the solution that works best for your needs. If the design changes, the electrical system automatically adapts to any modifications in the geometry, electrical loads, or other requirements.

How it Works

To get a better understanding of how Augmenta Electrical works, let’s look at a detailed example of how it is used to design the electrical raceway systems in a sample building. The inputs to the application — which is installed in Revit as a plugin and is code-named Acorn — are the overall geometry of the building, the locations of the electrical rooms on every floor of the building, and the electrical junction boxes in each of the rooms (Figure 3). These locations are determined by the electrical engineer and modeled in Revit, just as they would in a traditional non-augmented design process.

Additional settings and requirements related to the electrical design are typically specified in a detailed wire schedule, also known as a circuit schedule. This is created by the electrical engineer for the project and specifies each of the to-from connections between the central distribution panel, the power panels, and the junction boxes. Additionally, the loads for each connection are also specified (Figure 4).

As the final set of inputs to Augmenta Electrical, you need to let the system know where it can or cannot route in the project. This is done by placing routing boxes in Revit using the Augmenta plug-in. As shown in Figure 5, there are three different types of routing boxes: Must Route, Can Route, and Keep Out. The decision on which routing box to place at which location is typically made by the electrical engineer or contractor working with the structural engineer. They can work together to determine which element can be punched through so that it will not cause a structural issue in the design.

Now that all the inputs have been defined, Augmenta Electrical is launched from the plugin using the “Launch Acorn” tool. Once the user initiates the generation process, the solution transmits all the inputs and requirements to the cloud, where they are processed to generate multiple possible solutions for the detailed design of the system. Each solution is optimized and fully detailed, with a specific cost and schedule evaluation (Figure 6). To make it easier to decide on which solution to select, all the solutions are presented in the form of a graph showing their cost and construction time, allowing the tradeoffs to be clearly seen. This was shown earlier in Figure 2.

It is also possible to create multiple what-if different scenarios in Augmenta capturing alternate inputs such as the location of an electrical room, the use of factory elbows, the use of bent conduits, etc. These are called “studies,” and as shown in Figure 7, the solutions generated by Augmenta for each study are shown in the graph in a distinct color.

Within each study, users specify a detailed list of design rules and settings that are used by Augmenta to generate the solutions for that study. As shown in Figure 8, there is an extensive set of rules for a study capturing the design of parts, grouping, and routing, which can be customized as required. This is where company standards and best practices as well as any national, local, or client-specific standards can be incorporated. The wire schedule shown in Figure 4 is also referenced here to capture the specifications of the system.  Since all the designs in a study are generated using these rules, they will always be accurate, spec-compliant, and code-compliant, including the sizing and the selection of the conductors for the connections and the grouping of wires in the conduits and raceways.

After exploring the different design solutions generated by Augmenta, you can select a solution to export to Revit, and it will now appear in the project, as shown in Figure 9. All the different components of the solution are automatically created in Revit using actual manufacturer-specific parts from the Revit library. If any modifications to the design are required, they can be made using the regular Revit tools.

The Use of AI

What makes Augmenta’s rule-based generative design system different from other rule-based solutions that can automate the design of building or infrastructure elements — such as site design tools in SITEOPS, automated house designs in Bluethink House Designer, smart infrastructure design tools in Autodesk Infraworks, automated stair and railing tools in Archicad, automatic reinforcement of structural elements in Allplan, etc. — is the use of machine learning to manage and guide the rule-based component of the application. Thus, the system is not static but dynamic, as it is constantly learning from new data sets of electrical designs that are used to train it. It helps Augmenta to, for example, capture unwritten rules that go beyond best practices and that are preferred by a particular client or design firm, which is something a traditional rule-based design system cannot do. Augmenta is thus a hybrid system that combines the continuous improvement of machine learning with the determinism of a rule-based system.

With regard to the data set of electrical designs that are used to train Augmenta, there is no real-world data set yet that is large enough to be meaningful, so Augmenta uses a synthetic data set created from its own rule-based system. It can create as many correct designs as are required to train the machine language component of its system. This is common in AI applications where there is not enough real-world data, such as in self-driving cars, for which 90% of the data used for training is synthetic and comes from a driving simulator. This is because there is not enough data available from real-world driving — it is too expensive to gather, and it takes too long to create. AI systems need data that is meaningful, fully correct, accurately labelled, and at a certain level of complexity to be useful for training, which is why an application like Augmenta has created its own data.


I found Augmenta a very promising application that is putting the cutting-edge technology of generative AI — which we have been hearing so much about — to actual use in the AEC industry by automatically generating detailed solutions for specific design problems that would otherwise need to be created manually. By plugging the technology into Revit, a design and documentation application that is routinely used in AEC, Augmenta can become a mainstream design aide rather than being relegated to an esoteric, aspirational tool. If the goal of technology is to reduce the grunt-work humans need to do, Augmenta seems to have hit the nail on the head for the AEC industry.

About the Author

Lachmi Khemlani is founder and editor of AECbytes. She has a Ph.D. in Architecture from UC Berkeley, specializing in intelligent building modeling, and consults and writes on AEC technology. She can be reached at


Have comments or feedback on this article? Visit its AECbytes blog posting to share them with other readers or see what others have to say.

AECbytes content should not be reproduced on any other website, blog, print publication, or newsletter without permission.

Related Articles

AI in AEC Updates, 2022

This article explores the application of AI (Artificial Intelligence) in tools across each of the three main processes in AEC: Design, Construction, and Operations/FM. Some of the applications discussed are TestFit, Reconstruct, and ALICE Technologies.

Why Isn't There a Smarter BIM Tool for Building Design, Yet?

This article questions why we still don’t have a smart building design tool which does not require us to painstakingly model every detail in our buildings, but can automatically create much of it from a conceptual sketch using a rule-based expert system.

Bluethink House Designer: Automating the Re-use of Design Knowledge

This article from 2009 explores the rule-based design technology of Bluethink House Designer, which was developed for one of the largest home building companies in Norway. It no longer exists but is a great example of advanced rule-based design in AEC.

SITEOPS: Applying Optimization Technology to Site Design

This article looks at the application of evolutionary computing technology -- which uses optimization to find the most effective solutions to a problem given a set of constraints -- to the field of site design and land development in SITEOPS.