Generative AI is quickly becoming a transformative force in construction and engineering design. By leveraging existing engineering knowledge and judgement capital, it has the potential to revolutionize how we design everything from skyscrapers to bridges.
Firms' ability to capture and utilize historical data and information within their human capital will differentiate their capacity to enhance their projects and workforce with generative AI. So, firms must understand better and address how they capture and collate tactic knowledge.
This article explores the impact of generative AI on construction design and the pivotal role of CalcTree, a platform that allows designers to harness years of technical expertise, design principles, and empirical knowledge.

The Untapped Wealth of Tacit Engineering Knowledge

When we speak of existing engineering knowledge, we often fail to distinguish between explicit knowledge, readily available in documents, reports, and databases, and tacit knowledge, also known as 'engineering judgement'. This tacit knowledge, expertise and know-how built up over years of practice, are stored in the minds of experienced engineers. It forms the primary competitive edge for consulting and design firms in particular, that's why these firms so fiercely compete for talent. But due to its nature, such information is hard to access and challenging to structure in a way that a machine can interpret. However, doing so is key to unlocking genuinely transformative design possibilities with data-driven AIs.
One of the most critical prerequisites for leveraging AI in engineering design is capturing and organising tacit knowledge alongside empirical data and explicit information. AI can only learn from and generate new designs using the data it's been trained on. Therefore, unrecorded expertise or undocumented design cannot be leveraged by an AI that needs that information to generate new ideas. So capturing, organizing, and making tacit knowledge explicit is essential in AI-driven engineering design.
Generative AI
This is where CalcTree, a calculation and design management platform, comes in. CalcTree helps bridge the gap between tacit knowledge and the data sets required by AI models. It does this by helping to capture and structure typically unstructured project information, like engineering calculations, design principles and notes and boundary conditions, and process knowledge. CalcTree integrates seamlessly with everyday tools such as Excel, Python scripts, reports and simulation models. It enables users to create and collaborate on templates, making the data within these tools accessible and interpretable by systems designed for AI training.
CalcTree is also designed to integrate seamlessly with other tools via API, serving as the perfect backend for engineering and construction firms exploring generative AI applications. Whether the tacit knowledge resides in a different design tool or database, CalcTree has visibility of it.

The Future of Engineering Design with AI

Let's take a closer look at the future of engineering design with AI to understand the importance of knowledge capture in more detail. Specifically, how the advent of bespoke AI 'co-pilots' might influence how workers do their jobs, and a company's overall productivity. Let's consider the development and implementation of such co-pilots using Mckinsey's '3 horizons innovation' framework.
Mckinsey's '3 horizons innovation' framework

Horizon 1: Knowledge Assistants for 'first drafts' and Ideation

Horizon one sees Knowledge Assistants predominantly used for ideation, composition, and text improvement within engineering and construction settings. AI models at this stage are task-specific, and the accessibility and bandwidth of a particular user restrict their interaction with corporate data. These assistants can help with straightforward tasks, such as generating new design ideas or providing first cuts of code and the like.
However, they still require much human input to be valuable. Additionally, as pure large language models (LLMs) they're very limited in their computational ability, so they have limited value, in particular in the computationally heavy field of engineering.
As well as this, at horizon one, security and privacy arrangements between 3rd party LLMs and secure and private company operations and projects are not established. As such, many companies are restricting their workforce's usage of tools like chatGPT and Bard.
We can already see the advent of generalised versions of these assistants in the market, Bard, chatGPT etc. But we expect to see more sector-specific versions of these tools, perhaps trained on more specific industry data or with better computational abilities, enter the market over the next 6-18 months.

Horizon 2: Enhanced Interaction with Private Data

As privacy and security arrangements are established in firms, the second horizon will see the introduction of AI co-pilots with the ability to interact with private data. This means these AI entities will have access to explicit company and project-specific data, giving way to tools that can generate ideas and drafts that consider such information.
In construction and engineering, at this horizon, we expect to see AI tooling that can reach into CAD, BIM, simulation software and regulatory data. Thereby able to provide insights and user assistance specifically tailored to context.
Some startups are already releasing betas and prototypes of these concepts, and it won't be long until the likes of Microsoft, and Google integrate such AI into their product suites as well.
The challenge for the construction and engineering design sectors is that project data is highly fragmented and typically spans dozens of data formats across isolated environments. So reaching horizon 2 in our sector will require broader and agnostic data management tools like CalcTree to reach into and 'flatten' data sets across tooling.

Horizon 3: AI Co-Pilots as Active Contributors

The third horizon would see Knowledge Assistants that can interpret data and then execute tasks on behalf of designers. Through interaction with data management systems and tools, AI co-pilots could start to perform actions such as producing and running simulation models and then applying computational rules and engineering judgement to designs. This would work to overcome the computational limitation of LLMs.
Reaching this stage will require addressing challenges related to data fragmentation and the diversity of tooling used in design processes. As well as training AI on collated tacit knowledge, as well as explicit contextual information.
Wide-scale deployment may be slower in specific sub-sectors with more design edge cases and niche tooling. So adoption of agnostic data management tools and smart usage of AI interpreters to process, organise and drive niche software will be critical.
Across all these horizons, the key to success and the increasing capability of AI co-pilots lies in the effectiveness of information collection and collation, as well as 'powering-up' general AI (LLMs in this case) with access to third-party tools and data sets.
This is particularly true for construction and engineering, where data is often spread across different platforms and systems. Platforms like CalcTree play a vital role in this regard, helping to capture and structure both explicit and tacit knowledge that AI can harness across all three horizons, with increasing importance at each level.
Integrations with critical tried and tested industry software like Revit, Tekla and simulation software - which is also possible in CalcTree, will also pave the way for horizon 2 and 3-level AI assistants.

Challenges and Future Perspectives

While the potential benefits are enormous, it's also crucial to consider the challenges and implications of generative AI. Issues around data privacy, algorithm bias, and the need for human oversight must be addressed. Future research and regulatory measures should focus on creating an ethical, inclusive AI framework that serves the needs of all stakeholders.
Additionally, the collection and collation of tacit knowledge, as opposed to just explicit knowledge, is key to unlocking truly powerful AIs which can apply engineering judgement to complex situations.
Generative AI is poised to bring a paradigm shift to construction and engineering design, unlocking new possibilities and better leveraging our rich engineering knowledge. The future will see AI not as a replacement but as a partner in design, working alongside human engineers to shape the cities and landscapes of tomorrow.