Using AI for document mining a big win for construction sector

15 February 2024 Consultancy.eu 4 min. read

By better managing their documents, companies in the construction sector have a huge opportunity to become more successful in their projects.

Leveraging AI to analyze and ‘mine’ important information from large document troves is a powerful way to cut through the noise and save time in the hectic and deadline-driven environment of building projects. This is according to Herman Hoekman, an expert at Dutch consultancy firm Semmtech.

To find out more about the topic, we asked Hoekman five questions on the need for document management and its key benefits for builders and constructors.

Using AI for document mining a big win for construction sector

Herman Hoekman is practice leader for Architecture, Engineering, and Construction at Semmtech

For some background: how dependent are construction groups on the use of documents, and what integration and process challenges do they face in bringing all this information together into integrated reporting?

Every organization within the construction industry relies extensively on (physical) documents for their daily operations, both internally as well as in their interactions with other organizations.

As a result, during the various projects’ phases a substantial amount of time is spent on searching, gathering, and extracting relevant information. This information is then manually transformed into a suitable format and structure to facilitate the workflow. Then, there needs to be an effort to deliver the document to its intended recipient. This iterative process comes with a significant administrative workload.

In order to streamline this process and capitalize on the value of data, it is essential to ‘mine’ these documents and extract the valuable information locked within them. Just like a physical resource, it must be extracted and turned into something useful. This process involves reading the text, identifying relevant information, and transforming it into a structured format.

What are some examples of the kind of documents that are most suitable for this mining and aggregation process?

Many different types of documents, often PDFs, are required at every stage of a construction project and are shared across different stakeholders. Several engineers, architects, and construction companies often work on the same projects and share documents back and forth throughout the various processes. That can be time consuming and can cause confusion.

By turning these documents – for example, standards and requirements, guidelines, handbooks, contracts – into structured information, the sharing and exchange of information is much more efficient.

How does this work, and can it be fully automated?

After an initial analysis of what type of documents they are dealing with and understanding the various use cases, the tech-driven mining process makes use of AI and NLP (natural language processing).

The required AI/NLP technology must be configured to autonomously process the documents. First, this involves setting up the necessary AI/NLP pipeline(s) for efficient and accurate data extraction and processing. Second, an iterative process of training, testing, and implementation is essential. This ensures that the technology is refined and optimized for seamless document processing.

Once successful, the system can independently handle document automation and extract the necessary information, streamlining the entire process.

Is it only good news, or are there also some risks involved with the use of AI?

Although AI tools can be immensely powerful and make processes more efficient, there are indeed still some risks involved. Sometimes AI can get things wrong because – though powerful and accurate as it is – machine learning is not perfect. User input is key for improving the tools.

Therefore the end-users play a pivotal role by offering valuable feedback, fresh insights, and adjustments accompanied by providing example training data. This input then serves as a catalyst for continuous improvement, enhancing the overall capabilities of the AI/NLP system. Without a feedback-loop into the trained datasets, the learning process can stagnate, hindering the ongoing enhancement of AI/NLP capabilities.

With technological advancement becoming a higher priority for construction companies, what is necessary to move forward?

We have seen first-hand that the construction industry in particular has a lot to gain from the wide range of tools that use AI and machine learning to save time and resources. These new and emerging tools have applications that can help at the design stage, support in mitigating risk at the building stage, and with measuring progress throughout the building process.

At Semmtech we know the construction and infrastructure industry, and we have learned that each use case demands a (slightly) specific AI/NLP pipeline configuration, and can quickly support in the steps required to set up this capability. We typically start small to prove the technology on a single use case, and scale from there.