Reducing 160 Work Hours to 30 Seconds: How AI's Advancing Sustainable, Protected Housing with 30% Carbon Footprint Reduction

JOSE VALVERDE: Reducing 160 working hours to just 30 seconds. Hello, everyone. I am Jose Valverde. For me and the team, it's a real pleasure to be part of this amazing conference.

I am the BIM director at Via Agora. And with my colleagues, Carlos and Ana, we'll be talking about BIM, which helps us to work with digital complex BIM models for developing projects, manufacturing, which enable us to industrialize and standardize construction in factories and artificial intelligence, which can optimize processes, forecast problems, and automate manual tasks to mitigate errors. In these sessions, we will explore how these technologies are not only transforming individual projects, but also reshaping the whole industry.

So today's session is about how we have reduced 160 hours of manual work to just 30 seconds using AI through BIM procedures for developing industrialized facade models. Thanks to the processes we have implemented, we decreased our material wastage from 40% to less than 9% thanks to our optimized solution. Currently, our automated design panel resolves 81 of our facade panel with a scope to further improve.

Automation allows us to reduce human error per project to less than 4% down from 60% when done manually. And finally, by integrating BEAM and AI, we have improved factory production by 850%. Quite impressive, right?

So how do we do that? Well, let me start from the top. As I said, today, we will discuss the integration of BIM, manufacturing, and AI. This has been made possible through the collaboration of three specialized Spanish companies, Via Agora, Lignum Tech, and Nidus Lab.

So here we are, Via Agora, the company that I work for, is not just a company. It is a corporation that has many companies that form part of them. Via Agora Corporation is made up of six companies focused on sustainable development. The real estate developer designs and builds residential projects, while Lignum Tech is a specialized company that industrializes construction systems and manages sustainable forests through Lignum Tech Forest, another company that is part of the corporation.

Our construction company, Via Agora Gestion de Proyectos, builds homes for Via Agora projects. And the Gomez-Pintado Foundation ensures the company meets its economic, social, and environmental sustainability goals. We also manage and produce award-winning premium extra-virgin olive oil with Finca La Pentazuela.

Moving on to construction, let's take a look at some of our facts and figures. Via Agora was founded in 2007, and as a company that developed real estate projects under the name of Via Célere with operations in Spain, Bulgaria, Poland, and Brazil. In this 15 years, we have sold over 7,000 build-to-sell homes in these four countries. Currently, we have over 800 build-to-rent properties and a workforce of over 270. 30% of our workforce are under 35, bringing in young buyers, making it easier to implement new technologies and the work modes.

As I was telling you, I am the BIM Director of the Real Estate Developer. What we basically do is to develop projects and build homes based on our three pillars, technology, manufacturing, and sustainability. We try to ensure to use as much technology as possible-- mainly, it's from Autodesk-- to make our work efficient, more accurate, and cost-effective. We make housing more accessible to society, especially for the next gen.

But the building sector started off in a traditional format, where the use of technology was not as widespread as in the engineering or automobile industry. We started from here, from a manual process. I think that it's important to know the origin of construction, to understand change, and the important role of technology. The construction is started in a social context, where many people develop project manually without coordination. The coordination and critical decisions that had that important economic impact of the business were left in the hands of the worksite managers and foremen.

We can see players working on a project that undergoes several changes run a risk of lack of coordination. Especially when the project duration was longer than it is today, manual project development ran a huge risk of human error.

Now we can see several technicians drawing plants by hand. They are a great professional, of course, but lack of coordination makes the project information inconsistent.

This is why the use of technology is an indisputable core of our projects. We use BIM to develop precise and accurate projects. In the image that we are seeing, it's a traditional project without BIM. There are many people, many tasks, a lot of information, lack of coordination. It's a hard system to manage. They are wasting valuable times trying to find information when they need it most.

On the other hand, here we have a BIM project. We use BIM to cater added value to our projects. It allows us to manage information precisely. With BIM, we can click on elements to see their properties and find information readily. It allows us to standardize and repeat construction systems and operations to save time and cut costs.

So this is what we do, create digital models that clearly represent the project to be built later. That's why whenever we develop projects, we ensure that the virtual designs match the final construction. The main intention is not just to create nice-looking BIM models. Our primary goal is to create BIM models with sufficient level of development, to ensure coordination between the developer and the manufacturer. We use BIM to anticipate any issues that may come on site, to solve them in their early stages at our office where the change management doesn't have a direct impact on costs.

Though at Via Agora, we face ambitious challenges that require a great technological input. If we want to be economical and sustainable, we can't keep working manually with lack of coordination. We need to optimize our processes and adapt them to currently reality.

Our goal today is to understand the process from the design to manufacturing on all the conditions that affect this process. We use BIM methodology when we start developing a new building design. Our architecture team designs a project that integrates all the construction disciplines, architectural, MEP, and structure. Once the project is defined, the specific manufacturing models are created, so in the geometry requirements for Lignum Tech. These models are coordinated with the architecture of the design to generate detailed engineering documents from validated BIM models. Finally, once the detailed engineering documents are approved by Lignum Tech, manufacturing of these industrialized systems begins in our factories.

This workflows follows the step-by-step process where we constantly audit the results that integrates all the steps and ensures that no information is lost. For that, in our kind of environment, we can't not apply the traditional workflows, where the information flows randomly among the players and projects suffer constant and uncontrolled changes.

With these BIM workflows, we centralize information in one place that everyone can access, ensuring traceability and collaboration among all players. Our tool for doing that is Autodesk Construction Cloud. It's a common data environment that ensures access to updated information, essential for any collaborative projects.

So process makes the difference. And how do we obtain projects that are valid for manufacturing at Via Agora? Well, we start with the project design. How we develop the projects, ensuring they meet Via Agora requirements and criteria. Next, we conduct a technical review to ensure that the developed project meets the manufacturing criteria. Following that, we have our collaboration project, where the architectural project and the manufacturing project are developed together with both updated at the same time. Well, now we audit the final construction document for any users to make the final corrections. And finally, we have a validated project, ready for manufacturing and construction.

And how Autodesk helps us in this process-- well, we design all of the projects with Autodesk Revit, and we apply Autodesk format for general and detailed perspective. We coordinate the different disciplines using Navisworks Manage to ensure 3D coordination of model. ACC Docs assures that all the players have access to updated project information and maintain document traceability. We manage project users using Usage Tool from Autodesk Docs, of the ACC Docs, ensuring effective communication flows. And finally, we publish validated version and integrate a QR code on each plan linking directly to ACC with access to the latest 3D version of the BIM model.

So BIM is the way to go.

Our path, our goal is do a standardized project and architectural design for industrialization. So our process never start with a 2D design. We avoid creating 2D design like this.

We directly develop projects with a purely BIM approach, creating a digital preconstruction to real estate, focusing on the value of information. Our goal is not just to create our pretty-looking 3D models.

We use BIM for-- well, BIM allows us to manage and supervise the flow areas of our projects. So developer, our main business is based on selling floor areas, their square feet. That's why developing BIM models with the precise detail to build correctly is paramount.

BIM also keeps cost control in check. Our business model must be aligned with the project design. For this modeling, how it will be built is essential to always keep the project parameters under control.

And the procedure, 3D coordination is critical. If a change in architectural design affects the industrialized components, we need to be able to identify those changes quickly and accurately.

So ensuring coordination that our companies work in harmony and collaboration of different BIM models, it's paramount for industrial success.

And now let me explain how we apply Autodesk Forma when we start a new project design. We start a design by modeling generalized spaces and volumes that help us obtain information. Here we can view the daylight potential of the building designs we have created, for instance. We can also view the hours of sunlight that the key areas like pool area or building facade has. And also, Autodesk Forma helps us to predict and analyze the critical forces of wind loads and incidents upon our building.

And now we come moving on the modeling process, where we coordinate the BIM model structure with the BIM models that correspond to the MEP disciplines in the building project in order to be able to reduce 160 working hours to just 30 seconds. So it's essential to follow certain standardization criteria that affects the design. These criteria are the core decisions of our future generative AI application.

So by following Lignum Tech recommendations, we design projects where one floor of the building is repeated as many times as possible. So we design one floor, and we have already obtained at least 70% of our building. And finally, with this condition, we design our project with a level of development like that.

This is a real image of the project already in the construction phase, that this looks like the BIM model.

And now let's thinking about the outside the box. The goal in this session is on everything discussed earlier, is to industrialize and save time. Together, we need to standardize, innovate, and think beyond.

At Via Agora, we aim to reducing construction costs, build faster and way more sustainably. That's why we don't hire people for manual tasks. We exploit technology and new solutions to the maximum.

That's the reason why Lignum Tech formed part of Via Agora Corporation. Let me tell you a little bit about what we do at Lignum Tech.

[VIDEO PLAYBACK]

[MUSIC PLAYING]

[VOCALIZING]

[END PLAYBACK]

Beautiful. This is all the processes. So once I was telling you that Lignum Tech industrially manufacturing system primarily produces facades and bath pods. We currently have five plants in operation, and we will be adding two more in the upcoming months. Our annual facade production capacity is in the order of 2,000,000 square feet and over 3,000 bath-pod units. Lignum Tech has invested over $12 million US, and we count on our workforce around 100 professional technicians.

Also, this session is mainly about reducing time, infrastructure engineering. I will share to some interesting insights about our industrialized bath pods, our bathroom models that we call bath pods. We manufacture fully equipped bath pods in our factory and then install them on site once the structure is completed.

To a successfully industrialization, we must to work hand in hand with the architectural team, identify the existing typologies, like a bath pod, recommend any modifications to standardize and reduce the human number of typologies, and develop the fully coordinated manufacturing project with the architectural design. To achieve this, we collaborate with the architectural team to only design a building that can be industrialized. So once the project design is defined, we identify these typologies of bath pods that best fits the project.

To do this, we develop a specific BIM model for our bath pod typologies that we will later use for manufacturing.

Then we start with a basic design, where we only define the general geometry of the bath pod.

And once the project design has been validated, we develop feasibility models where we include more information to ensure effective coordination with the building system engineering.

And thanks to ACC version control and management, it enables us to maintain a complete oversight of changes and ensure coordination between models.

This process enables us to develop detailed models, allowing us to carry out detailed engineering for manufacturing the bath pods in a semi-automated and error-free manner. This is how we develop models that our plant operators use to construct the bath pods.

We do BIM models, but also a 2D drawing's information complements that 3D models. These drawings, thanks to the application that complements ACC, we include a QR code on each drawing, and with this QR code, directly links the 2D drawing with the 3D BIM models located in the ACC to ensure the latest version of the bath pod is available.

And finally, this is the result of our bath pods once they have been constructed and transported to the site.

We deliver the bath pods fully finished with our installations. Only the connection to the main networks and the drywall finishing of the models are required.

And this is the interior result of our industrialized bath pods. We achieve the same quality as a traditional system, but with a significant reduction in the execution time.

So we believe in an industrialization to be more sustainable and efficient. And we know that we live in a context where there is less available labor each time, and the construction costs are constantly rising. That's why we need to find solutions. As we have seen, we have built five automated production plants where we produce industrialized systems, reducing physical labor in the manufacturing process.

Our factory is divided into different production areas, each specialized in a specific task. Here we have the batten mounting station, for instance.

With the help of robotic arms, we build our industrialized system in an almost fully automated way, just like in a car factory, for instance.

The machinery receives the information from the BIM models and performs tasks at different stations. In each image, for instance, the robotic arm is securing the wooden battens that form the structure of our facade model.

So we carry out our tasks without printing out drawings or plans in a paperless environment. Our team works in the real time with BIM models.

And they only have to verify that the BIM models match the final machined construction.

Once the work is completed, we prepare the facade models for their final finishing.

That is this. This is a facade model with the final finishings done.

And then when the finished models are finished, they are being transported from our plants to our construction site.

And here at the construction site, all we need to do is to mount the models onto the structure. This is one of our sites in the mounting process.

More images about our process. Here we have a completed structure at the right. Next is the mounted facade. So thanks to our process models, the time has been reduced in execution time.

This is amazing, but these types of projects are a real challenge, and they are constantly require new ways of working, continuous learning, and new technologies.

So let me show you an example for a project where we had an existing building in the middle of our site. We called in the land surveyor to provide us with this data to help us to coordinate and developed our project. But this is not enough for us. We need to work with more details.

So what we did was create a point cloud. We scanned the existing building to ensure project design was accurate and to avoid any errors in the facade design. Then working together with the point cloud, we were able to successfully coordinate the project and industrialize with confidence.

So we managed to reduce 160 hours to 30 seconds, but it was not as simple as that. To get where we are during our learning curve, we discovered and developed efficient techniques.

We started by modeling Revit families with a basic level of detail for initial designs, which we later replaced with more detailed ones as the project progressed.

These Revit families had an excessive information, requiring extensive maintenance and not easily to adapt them to the project design.

Before applying AI, we create ad hoc BIM models for facades, where we manually modeled each model, as the image that you are seeing. Then we coordinate these models with the architectural Revit model. We use for that Revit ACC Model Coordination or Navisworks Manage, for instance. This depends on the size of the projects.

But any project change means a huge amount of work for us. So we decided to develop new solutions to design our project quickly with less effort and more precision because our goal has always been to move from this to this in the shortest time and with as much automation as possible.

So now let's talk about the industrialized system, and my colleague Carlos will explain what are our industrialized systems. So, Carlos?

CARLOS LOPEZ: So Jose, thank you for awesome introduction. And my name is Carlos López and I have the pleasure of working as a Technical Director at Lignum Tech, a company that is part of the Via Agora Corporation. Today, I'd like to talk to you about industrialized system design process, which is not only timely, but essential, as we push forward in this ever-changing landscape. Whether you're a seasoned veteran or just getting started in this field, my goal is to serve fresh insights, new ideas, and maybe even some questions we hadn't considered before. Thank you for having me. So let's dive in.

As previously mentioned, Lignum Tech is company whose main activity is the manufacture of industrialized systems for high-rise residential construction. We've developed an industrialized facade system using a light timber frame with a self-supporting structural system made from timber. This is not only an innovation from a technical standpoint, but it also offers high performance. What makes this system stand out is its adaptability to both new constructions and restoration projects. It's efficient, sustainable, and integrates perfectly with modern architectural designs.

Next, let's talk about staircases. Our formwork system is made from high-density insulated materials that are coated with synthetic resin. This system is specifically designed for cast-in-place concrete structures and comes prefitted with reinforcement.

Moving onto balconies, we've created the steel-structure components that are designed to complement our facade systems. One of the key innovations here is that they don't require scaffoldings. Instead, they are anchored to pre-installed points, making them much faster and safer to install.

And finally, we come to bath pods. These are 3D industrialized bathroom components that are fully manufactured offsite. They are manufactured with dynamic layouts, a self-supporting structural base, a perimeter partitioning system, and top-quality installations for easy connection to general networks. They are delivered fully finished, which means installation on the site is quick, saving both time and labor.

To develop the subject, we are going to focus on our facade system, which, due to its multiple interrelated layers in the construction system that may require more optimization criteria, material savings, collisions between fasteners, safety zones, et cetera. In the case of the curtain wall facade, the panels are made with a light wooden framework that can range from 3.5 to 5.5 inches thick. Once a structure is calculated for its self-supporting use, it is covered with closing panels of various materials, such as bitumen boards or oriented strand boards, a veritable waterproofing layer, and finally, vertical and horizontal timber battens are placed, dependent of the finishing panel to be used.

In the case of the structural load-bearing facade, the light-wood framing does not pass in front of the floor slabs. Instead, it is located between them. The layers may be the same, but the thicknesses will vary from 5.5 to 7.5 inches, in some cases using sophisticated wood products such as LSL or [? GLBL ?] instead of some timber.

I do not know the current state of industrialization in the United States. I'm aware that this is far ahead of Spain, where we are currently taking our first steps in the sector. And I suppose you have faced the same challenges that we are facing now. But the question is, are we aware of what implementation of a new robotic industrialized line means in numbers?

The main problem arises when we look at the level of indirect cost related to personnel. For engineering department with around eight employees, it results in a manageable quantities of 1,500 square feet of new detailed engineering per week, using the additional design methods and the specialized software currently available on the market. To put this into perspective, we would need to multiply that workforce by 11 to match out the output capacity of the robotic facade line.

The ideal moment to introduce industrialization in a project is during the initial planning and design phases. This allows processes and solutions to be adapted to the project needs from the outset, maximizing the benefits of industrialization.

Ideally, it should be implemented during the conceptual design phase, as this facilitates the integration of prefabricated or robotic construction systems, optimizing time, cost and quality. Additionally, in projects with tight deadlines or when precision and repeatability are crucial, industrialization becomes an ideal option.

So highlighting the main advantages of industrialization in building projects, we find reduction of lead times or lifting problems, resolves lack of the storage space, cope with lack of skilled workers, meet the need for high quality, optimal for job complexity, reduces the incidence of accidents, and improves recycling materials and reduces waste.

During the initial stages, the panels were designed using specialized 2D and 3D software, resulting in a high error rate, extended design times, and repetition of many actions that could potentially be automated in the future.

So each panel had to be done with each layer on a separate plan. There was no coordinated information between the plans, and there was no design automation or associated information to the model.

In the second phase, new BIM software and dynamic blocks are introduced to automate the design process, coordinating information and reducing design times and error rates. With implementation of new software, detail engineering is created from zero, design automation begins, and the standardization of components is introduced through dynamic blocks. Repetitive actions are almost eliminated. Information is coordinated within the model, and change management becomes much more effective.

This allows us to start from a coordinated model that is managed in real time and manufacture industrialized models that are delivered to the site as exact replicas of the model design and managed in the previous months.

We create BIM families and implement element module solutions, once again aligned with the objectives of speed, precision, and sustainability.

The creation of BIM families led to the elimination of dynamic blocks. They come with associated information, such as cost, carbon footprint, and traceability. Discrepancy with the manufacturing base is minimal, and repetitive actions have been completely eliminated. Meanwhile, human error has been reduced by 80%.

The next step was to apply the standardization criteria to projects in the early design phases through the creation of catalogs.

Through the catalogs, we achieve detailed engineering and accurate project costs from an early stage. Indirect costs are reduced. And since the models are tested, error rates and economic deviations decrease exponentially.

Once the robotic line became operational, we had to implement a specialized software in machine language. But this resulted in a decrease in speed and precision, as well as an increase in design errors due to technicians having to learn a new software.

At that point, we needed to work through to BIM software applications for different project spaces, where the processes of standardization and automation have been taken to the limit. We now control the associated project information and manage change efficiently. But it is not.

If we tried to summarize the learning process through these phases, we can see that human error has been reduced, all processes have been automated, and the discrepancy between design and manufacturing time has decreased. However, as mentioned earlier, if our goal is to meet the production output of the fully operational robotic line, it is not enough in terms of speed, precision, optimization, and sustainability.

Once we have understood the layers that make up our system, we will simulate the design of a single panel with a couple of openings. The first thing we focused on is the screws' connection between the wooden slats and the top and bottom plates of the facade panel. Here we need to define maximum and minimum distances between the slats, tie pieces, and the distances from openings to the edge of the panel and vertical structural elements. Once the vertical elements of our light timber frame are defined, it's time to focus on the horizontal elements which are attached to the wooden slats with structural staples to ensure the adjustability.

The final is the definition of the connections between the structural elements of the light timber frame. We define the relative distances and position of the window joints, both vertically and horizontally in relation to the adjacent stats. The oriented strand boards are installed over the light timber frame with vertical screw lines along each of their sides, as well as the vertical axis. These lines must not coincide with any of the previous marked points and must adhere to the maximum distances from the panel edge and between the screw lines of the same panel. Then that set of lines and connection points between linear elements defines safe screw zones for the subsequent layers to be installed, vertical wooden battens, and finishing panels.

Now it's time to introduce the screw lines for the vertical wooden battens, which must be fixed to the light timber frame, passing through the boards and avoiding all the previous safety zones. They have maximum and minimum screw distances and must be arranged at regular intervals according to the finishing panels that will be installed over them.

We cannot forget the finishing panels which come with its own set of rules. Screw distances from the heads, maximum vertical and horizontal distances between different panels, and also screw lines within the same panel.

This is a very simplified demonstration of the effort required for a technician to design a facade panel. Technically, we've seen that with time and right decision-making, it can be done. However, we must remember that designing industrialized systems for a robotic production line involves considering several parameters, speed, design, error rate, sustainability based on material waste, and optimization of design parameters. Is it possible for a human mind to optimize all these decisions at the same time? Obviously not, at least not alone.

To explain this, let me introduce my colleague, Ana Lozano, who will explain AI design processes applied in the field of generative architecture.

ANA LOZANO: Thank you very much, Carlos, and thank you again, Jose, for your presentation. We got a lot of information there. And here I am willing to take a step back because, obviously, we had this bunch of restrictions that are going to apply on the project. But we need to learn, after all, how AI is going to help us on this process. So what we are building at Nidus is an idea to put AI of service in mostly optimization and automation, that it's going to lead us to design better and therefore to build better.

"The best way to predict the future is to create it," says Peter Drucker, also attributed to Abraham Lincoln and Alan Kay. So this is what we are doing here. We are trying to guess what tools could be of interest to completely revolutionize the real estate industry.

All right. What we call the real estate investment puzzle, as we've seen before, there's a big point applying into property investments that relies still to manual processes. And this is leading us to inefficiencies and missing opportunities and potential losses. And the technology today offers us an extreme and important possibility to make all these differences by turning manual and inefficient processes on processes that are assisted and driven by AI and other technologies.

If we take a look at three examples in which we could apply these kind of processes today, we are going to focus specifically on the tool we produce for these facade panels. But we can face these same problems across the whole industry, so not only in typical industrial manufacturers, with all the problems that Carlos presented to us and the challenges, but also everything related to create value out of a blank plot and even, though, for refurbishment for both developers and investors.

Just a quick reminder on what AI means because we tend to think that AI falls only under the layer of what we know now under the ChatGPT techniques. So this is one of the techniques that refers to conversion to large language models. But it's not the only way we can put artificial intelligence at service.

So we are going to focus on how the pure ability of any computational design or system to perform a task of typical human intelligence. They are going to be to the simpler and linear correlations to the more complex, including neural networks with multiple layers. What we are aiming to do here is to algorithmize a manual process, but to keep the human in the loop because, ultimately, it's a big difference between other systems in which we have data sets that are consistent and big enough, which is something that will not occur in the real estate industry.

So let's compare what was traditional thinking when we start with a plan or a design. We think of the problem. We search the opportunities. We produce a solution. And only after the whole process, we can evaluate and confirm or we need to start from scratch and iterate again.

What we intend to do using AI-driven thinking and producing generative AI specifically to serve this process is where iteration is an automatation of this thinking process. At the end of the day, the solution is something we can test several times and compare different scenarios to only keep the ones that are optimal.

As Lars Hesselgren has said before, Generative design is not about designing the building. It's specifically about designing the system that builds a building.

Let's come back to a building. So imagine a building like a bunch of LEGO pieces with no playbook. There are millions of options to assemble them, but we need to pin out the most optimal one. And what happens? 90% of the construction projects fail. They fail because of the delays, of the budget overruns, because they fail to meet the goals, or for a combination of the three. And this is mainly produced by, as we said before, the traditional process, which leads to uncertainty, lengthy processes, human error, inefficiencies, redundancies, inaccuracies, and at the end of the day, very, very high risks.

Just to remind that the tool we are using that we are presenting to you today could be applied in many other fields of the real estate industry. A use case, a real one, you can see of the background of the picture, the built building. So if we compare what would have been before Nidus, the traditional process, a team between four and six architects needed, a long time frame, a high rate of human error, a big number of repetitive tasks, not adding value to the process, and a low level of ability to recur and identify and iterate in different scenarios.

After applying Nidus, we could reduce the number of users and the timeline, the human error, and redundancy. But the most important is we could compare an unlimited number of scenarios, and we could increase the performance in number of units between 3% and 8%.

Let's focus now on the tool specifically built for Lignum Tech. We are going to present you a short video here of the whole process. So it's only a couple of times compared to the real speed. So we have this initial DWG format we need to resume and keep all the parameters that are going to impact on the facade design. Then there's a web app with a very intuitive interface. We just need to upload the very same DWG we produced, and the system is going to adjust the-- ask to the user to introduce the number of parameters that typically apply in the very detailed engineering design.

So we are going to test the model at different stages. First of all, it's going to be the size of the panels, taking under consideration the limitations, the size, logistics, and many other items my colleagues explained before me. You have on the page-- sorry for the Spanish. So far the version is only in Spanish. You have all the information in live that is going to be applied any time we produce these modifications on the parameters.

So the algorithm is going to be able to iterate a number of times and present the user with all the optimized solutions at each iteration. The user can choose different options. They can modify the panels. They can modify the finishing panels. They can act on so many parameters that, ultimately, the solution is not any more a black box, but mostly a Copilot and a design assistant that is going to allow the user as an expert user to be able to interact and behave and take the best choices within the use of the tool.

At the end of the day, when we include everything related to these engineering design, we can choose one specific panel by these numbers shown on the facade. And we can have all the information, very detailed information, even though, again, testing how the rest of the parameters are going to immediately apply and present with the best solution for the user. So the demo is just going to work on some of the different parameters that are ultimately going to change the design that you can see in the screen on the different colors. Then by using and modifying the parameters, the user can immediately act over all the detailed restrictions that my colleague showed before.

I'll let you just check on the demo. Here we are going to act on the separation between the different pieces. We are going to act on any of the decisions impacting on final model. When the user is satisfied with a solution, he will have the opportunity to obtain several outcomes. Some of them would still be in two dimensions, and some of them will be in three dimensions. We are going to check all of them.

So this would be the facade panel transferred from the original one in which we can cross-check all the dimensions that we obtained from the optimized model. This would be the three dimensions with all the elements interacting and, again, being able to be cross-checked by the professional expert user. And this is going to be another side of the very detailed panel. So at the end of the day, we can compare both.

This took us seven minutes, and the video has been here multiplied in speed by two, just to not to get you guys bored. Oh, yeah, so yes, the algorithm is going to be able to iterate 100,000 times a second, meaning that we would need eventually 100,000 engineers being able to produce a solution, an iteration, every second. It's still amazing.

Another success story based on the same similar tool as we saw before, another project that has already been built and, again, an opportunity for different verticals in the real estate industry to apply this kind of knowledge.

But we don't want to stop here. We are aiming to go beyond and to take the opportunity of AI to act as a catalyst between the different stakeholders and avoiding people to work on silos, but preferably offering this very powerful design-and-build tool in which all the tests that are conducted from a viability design can be compared to the systems in the industry in order to make everybody work in a collaborative way. We can also have here an opportunity to introduce circular economy and see how we could be using reutilized materials coming from other sites.

And as Arthur Clarke likes to say, "Any sufficiently advanced technology is indistinguishable from magic."

CARLOS LOPEZ: Thank you, Ana. Now let's break down the design process, comparing the traditional method with a generative AI approach. The initial phase involves identifying and defining the problems you want to solve. It requires research and understanding of the context, constraints, and user needs. Once the problem is defined, you explore existing solutions. This involves market research to see what's already available, assessing their strengths and weaknesses.

After identifying potential solutions, you design and create a prototype. This phase may involve iterations based on feedback and testing. Finally, you assess the effectiveness of your solution, looking for areas of improvement. These evaluations helps refine the product and can lead back to the design phase if needed.

The phases within the generative architectural design process are similar to the traditional process. It consists of the same phases. However, there is a shift in the order, fully exploiting the potential of generative design tools. We'll start with a problem to solve, but instead of searching for existing solutions, we need to think about the ideal parameters and characteristics for our design.

At this stage, the tool is able of generating millions of solutions in just minutes, extracting the most optimized ones from the databases. With this, the final step is to decide which of the proposed solutions best fits what we are looking for, prioritizing certain parameters over others. In essence, while the traditional design process is linear and sequential, the generative AI approach allows for a more iterative and data-informed process. This not only enhances creativity, but can also lead to more effective and tailored outcomes. The benefits of generative AI are speed, accuracy, innovation, and optimization. I'm going to come back to these benefits in a minute.

Generative design applications should not be misunderstood as software for a specific outcome, but rather as a tool. Once we understand the process, we need to optimize the conditions of our solution instead of seeking a single solution. We learn to work on not just one solution, but on millions within our databases. Let's take a look to the different prototypes developed by Nidus in collaboration with us.

The first prototype, we define the dimensions of each panel, the locations of openings, a structured access, and the distances between the union of panels. The goal was to optimize dimensions and begin standardizing the project using the vertical and horizontal structural axis and the location of the opponents.

Once we resolve the first floor, which was repeated throughout the building, we moved on to generating entire interconnected facades. We focused on repeating panel typologies and standardized the number of panels per floor, always prioritizing the alignment of vertical panel joints to allow for proper connection between them.

The next step was to introduce the optimization criteria of material waste. As we've seen in the simulation example, its panel can technically be resolved if enough resources are available. But that's not what we aim for with the application of generative design. We analyzed the main components of the system and focused on those that have the highest waste and were the most economically unfavorable, finishing panels and oriented strand boards.

This prototype is the first to incorporate the detailed engineering for each panel individually. Prior to that, the automated generation had to be done for each panel one by one. For the first time, it simultaneously integrates optimization criteria for distances, screw lines, safety zones, material waste, and sustainability, while automatically generating all the layers that make up its frame. The result is outstanding. Each panel generated by the tool took about 10 seconds compared to the six hours it might take a technician to design a single panel.

The final optimization introduced the simultaneous generation of all the panels for all the facades of the building. It reduced design time by 96% and design error by 35%. The system selects solutions with the fewest panels per floor and the least material waste, choosing the top 10 from millions of options. It then exports the entire project to the specialized software for the robotic assembly line in just minutes.

One of the main goals was to deal with the high cost of material waste due to design complexity, the set of rules to avoid collisions, and safety zones. We can see how this statistic has improved for its material based on the results of its prototype. Linear elements like battens have little room for improvement, but it's the 2D components and the structural elements, due to plate dimensions and design complexity, that have achieved the greatest waste reduction through the optimization of various criteria. Notably, the finishing plate saw a reduction in waste from 38% in the traditional design phase to only 6% in the final prototype.

In the same way, the optimization process translates into speed, and volume of detail engineering increased to 850%. 81 of all panels are automatically designed. The reduction in design errors that could cause shutdowns on the production line decreased to almost 4%.

To summarize the design development through the different stages, we have used the parameters of optimization and material waste percentage, design speed, precision, and reduction of human errors.

So for engineering department of eight employees that design 1,500 square feet of facade per week, we would have had to multiply that workforce by 11 to reach the 13,000 square feet output that the robotic line could produce. With the implementation of generative design applications, we have been able to increase design speed, automating 81% of the panels with a 4% error rate. This reduction leads to fewer factory shutdowns, which, translated to cost per shutdown, resulting in a profit multiplier by 20.

So what's next? We are currently designing with Nidus processes that can manage large quantities of facade and bath pods catalog models. Each of these models has been designed using the tool created in the initial phase. In the final stage, the program is able to identify matches between the imported project information and the catalogs, allowing us to receive project recommendations in a matter of minutes. These projects will be able to deal with small changes in an early project phase to generate a list of facade and bath pods typologies.