Integrating AI with BIM

What is AI?

Traditionally, developers programmed tools by imagining diverse use cases and customer needs (through Agile, Waterfall, or Hybrid approaches). However, they flipped the switch for AI solutions and began training machines to learn from a massive amount of data - billions of data in most cases (supervised or unsupervised).

Through Deep Learning, advanced machine algorithms can create new information from the vast amounts of data (texts, images, videos, and audio) used to train them. Thus, they can perform tasks typically requiring human intelligence, such as Natural Language Processing (NLP), Visual Analysis (via Computer Vision), or Arithmetic Calculations.

The primary difference between AI tools and traditional ones is that while the latter act strictly on their programming, AI tools can process and generate new information with the vast datasets used to create them - sometimes in unpredictable ways.

Nonetheless, present-day AI solutions are still 'Narrow AI' because they typically excel at one task type. ChatGPT, for example, cannot drive a car (yet), and Midjourney cannot perform arithmetic functions.

What is BIM?

Most BIM tools today are traditional tools programmed to carry out specific tasks. Where they run short, plugins, add-ins, and macros extend their capabilities. Using Revit as an example, there are countless plugins on the Autodesk App Store. Dynamo can stretch Revit's functionalities and help with Design Automation. Creating custom macros with C# is also possible. However, these are typically traditionally programmed just like the authoring tools.

So, how does AI come in?

Before we explore integrating AI with BIM, let's quickly examine the evolution of Digital Transformation in the AECO (Architecture, Engineering, Construction, and Operations) sector.

The Evolution of Digital Transformation in AECO

In the 1980s, CAD disrupted how AEC professionals communicated their designs with the emergence of tools like AutoCAD and MicroStation. Yet, before the 1990s ended, BIM began emerging, eventually accelerated by tools like ArchiCAD and Revit.

However, interoperability was one of the first challenges with the insurgence of several CAD and BIM tools from diverse software vendors. To tackle this (customer-facing) challenge, the software vendors agreed to standard formats for exchanging information between their tools, such as IFC.

Nonetheless, the file-based information-exchange approach still had many limitations.

So, as the internet and cloud services gradually matured, it wasn't surprising that the vendors sought better ways to connect project data through cloud solutions.

As the industry grappled with BIM evolution, we also witnessed the emergence of cloud-based CDE (Common Data Environment) solutions in the 2010s, thanks to platforms like Autodesk Construction Cloud, Procore, and many others. These cloud-based tools enabled project stakeholders to collaborate in real-time, but it quickly became evident that Data Exchange is more efficient on the cloud than through file formats.

In other words, CDE (customer collaboration) might be one of the first reasons software vendors migrated to the cloud, but arguably, the most pressing advantage today is Data Exchange. With solutions like Speckle and Autodesk Data Exchange, you have minimal need for IFC - these recent solutions focus on the data and can connect with multiple software from diverse vendors.

Unsurprisingly, most of the emerging tools in AECO are cloud-native, 'file-less', solutions. But again, why? Recall how developers build AI solutions - train a machine with a vast dataset.

I will summarise the evolution of digital transformation in the AECO sector as follows: we moved from CAD (which started in the 1980s) to BIM (which began in the 1990s), then to Connected BIM (which started in the 2010s). Ultimately, Connected BIM will resolve interoperability issues by adding a Data Layer above CAD and BIM tools while providing the granular data needed to advance the emerging next-gen AI tools.

Many AI tools are emerging daily, which can be overwhelming—it is easy to drown in the hype. So, we must discuss (organisational) implementation strategy before exploring specific AI BIM Use Cases.

Organisational Implementation of AI for AEC Firms

AI is one of those strong trends connecting every stakeholder in the AECO sector because it has a strong business case. Like BIM, three pillars must support every AI implementation in a firm: People, Process, and Technology.

It's easy to identify isolated use cases of AI and approach them ad hoc—today, Glyph Copilot in Revit, tomorrow Custom GPTs, the day after Swapp AI, and so much more. Such an approach can prove more expensive over time and still not add much value to a firm's bottom line—a case of doing AI for the sake of AI.

A more cohesive approach is for a firm to analyse its entire business process, technology stack, and talent and then identify areas where AI can increase efficiency—through task automation, augmenting human efforts, or better insights via Data Analytics.

1. It might require some organisational restructuring in terms of talent. Depending on the firm's nature, the adoption of AI might be centralised (around a hub) or decentralised (across several spokes).
2. It might entail employing some new people full-time or on contract - or even outsourcing some tasks.
3. It could also entail 'TAKING' off-the-shelf tools (e.g. Luccid AI), 'SHAPING' some integrations (e.g. creating a custom GPT with the firm's standards and documents), or even 'MAKING' a fresh solution with Machine Learning (if the firm has sufficient data and budget).

The analysis should focus on the change initiative's cost implications and projected ROI.

Potential cost implications include cloud expenses, prompt engineering costs, talent costs, operation costs, infrastructure costs, data security costs, and more.

The outcome will be a flexible Implementation Plan. No matter how vague the plan is, it's better than jumping on AI tools because of the hype. Over time, the firm gains more insights and updates the plan.

Now, let's explore some use cases and tools and how to integrate AI with BIM.

AI BIM Use Cases

These are some of the low-hanging fruits for integrating AI with Building Information Modelling:

1. Project Documentation: Documentation, Detailing, and Annotations are time-consuming and not-so-fun activities in every project. We can harness AI tools to augment or automate some of these tasks. Some promising tools in this category include Glyph Copilot from Evolvelab and Swapp.ai.
2. Generative Design: Generative Design promises to be a future approach to design because it helps define outcome criteria from the beginning of a project. Generative design algorithms can help designers analyse countless design options, simulate numerous scenarios, and conduct extensive analyses, thus basing their decisions on actionable data. Some promising tools with generative design features include Autodesk Forma, Digital Blue Foam, Hypar, Consigli, Giraffe, etc.
3. Code Compliance: AI tools can also help ensure that projects meet a jurisdiction's code requirements, allowing project teams to speed up the time between inception and final submissions. Some promising tools for code compliance are Struck.build, Luccid.ai, and Up.codes.
4. Model Generation: AI can help us automate the generation of models from PDF, CAD, Images, and Point Clouds. Some promising tools here include Bimify, Bimit, and Aurivus.
5. BIM Authoring: The next-gen AI (BIM Authoring) tools will be cloud-native, file-less solutions. Popular upcoming solutions include Infurnia, Snaptrude, and Qonic, as well as the tools listed under Generative Design.
6. Engineering: AI can be leveraged for diverse engineering use cases, from producing structural timer designs to analysing concrete (and more). Promising tools in this category include Daisy.ai, Augmenta.ai, and Structure.plus.
7. Construction and Analytics: AI can be helpful during the construction stage of projects - from quantification to document analysis to ConTech to Data Analytics, among others. Promising solutions in this category include Sparkel.ai, Document Crunch, Inspectmind.ai, Naska.ai, Conxai, Deep Space AI, and Autodesk Construction IQ (integrated into ACC).
8. Building Operations: BIM spans the entire life cycle of built assets, and so does AI. The benefits of great BIM implementation on a project manifest most during operations, and AI can enhance the benefits even more. Some promising tools in this category are PowerBIM, Autodesk Tandem, etc.
9. Business Operations: AI can improve AECO firms' day-to-day business processes. That is where assistants and LLMs (Large Language Models) like Microsoft Copilot, ChatGPT, Claude, Perplexity, Fireflies.ai, and the like come in.
10. Interoperability: With granular data coming from cloud-native, next-gen tools, interoperability will not be an issue. Today, solutions like Speckle and Autodesk Data Exchange connect multiple tools by extracting the data from various file formats.

Check out Stjepan Mikulic's AEC AI Hub for a more comprehensive list of AI tools in AEC. It currently features over a thousand tools and is continuously being updated. You can also join his upcoming Mastering AI in AEC course here. As the name suggests, it will set you in the right direction towards Mastering the applications of AI in AEC.

Barriers and Way-forward

The success of AI implementation and advancements in the AECO sector relies heavily on granular data. That encompasses project data (from customers) and context data like demographic, geographic, and climatic data, to name a few.

How much access do software vendors and developers have to train models that will yield accurate results?

For context, ChatGPT and similar LLMs scraped all publicly available data on the Internet (and more) to train their models.

The other side (of the coin) is that the AECO sector is highly regulated.

To what extent will regulations and standards impact the use of AI tools on projects?

For example, can we use AI to carry out seismic analysis during the detailed design stage of high-profile projects in the future? How do we explain to the regulators how we arrived at the results from a black box (the AI)?

Nonetheless, it's great to see the prominent software vendors signing agreements and partnerships to share data and open up their data ecosystems.

Let's see what AI has in store for the AECO sector in the coming years.

What do you think?

Register here for my upcoming free webinar with Zigurat: 'Integrating AI and BIM for Advanced Solutions'.

In addition to the points I have laid out in this article, I will dive into some tools to showcase how practitioners in the AECO sector can leverage AI to optimise and enhance BIM implementation in the industry.

PS: This article is not AI-generated; some images are AI-generated (indicated in captions).