Technology’s impact on the construction industry cannot be overstated. The industry has seen significant advancements impacting areas such as bidding/estimating, contract and insurance review and compliance, scheduling, safety and overall project and document management. The advent of smart phones, tablets and wearables has brought this technology out to those in the field, and allows data collection directly from their location, further increasing the value delivered.

Continued advancements in artificial intelligence and machine learning and the application of these principles to the construction industry are becoming recognized for what they are—powerful tools to reduce and address construction risks encountered before, during and after construction of a project. Understanding the technology, where to find it and how to use it are critical first steps to reaping the tremendous risk mitigation benefits.

What are “Artificial Intelligence” and “Machine Learning,” and are they the same thing? Artificial intelligence, often referred to simply as “AI,” is the broad idea of using machines (e.g., computers) to execute tasks otherwise requiring human review and operation. AI sometimes uses “Machine Learning” to automatically adapt and learn from the processing and review of prior data without express updating by human reviewers and programmers.

The application of this technology to reduce risk in the construction industry was greatly expanded by the advent of Natural Language Data Extraction, which allows AI platforms to identify and extract data from unstructured text in any format or order. Algorithms and platforms now exist to allow users to automatically identify, extract and categorize natural language from various sources such as contract documents, insurance policies, daily reports, change orders and RFI responses.

For example, imagine the power of a platform that can automatically identify design issues necessitating a change order from RFI responses, locate information in contract documents and insurance policies, or automatically identify and calculate weather-related delays from schedule updates and daily reports. Identifying and extracting issues and data from natural language displayed in any form is complicated and requires the appropriate AI platform and appropriately trained and developed algorithms.

One of the most prevalent and highly developed uses of AI in construction revolves around image recognition. Many companies have taken advantage of this type of AI and machine learning from simple use cases such as photo tagging, which is similar to running a Google search for images. The platform can identify all types of information from video and photographic images, often critical in tracking the status and progress of construction projects. More specifically, some companies have built predictive models to identify safety risks on projects to avoid potential incidents or accidents. This use case can provide tremendous value, especially now with the focus on safety measures such as the wearing of masks and social distancing.

Another area where AI and machine learning have impacted the construction industry relates to scheduling. Given the availability of historical scheduling data and its somewhat structured nature, several companies have been tackling the challenge of providing insights into project scheduling by developing models to optimize and predict schedule results based on historical data. This allows alternative schedule options to be evaluated instead of locking in the first one that works. While this technology is early in its development, it is anticipated to provide significant value and transparency to the construction industry and help reduce the risk of delays.

With many different platforms and applications now offering users in the construction industry the ability to unlock the power and “magic” of AI and machine learning to extract data, including from natural language, knowing how to evaluate platforms is important.

Accuracy is key, and AI technology is not perfect. Users want to find all the relevant information without anything being missed, but also do not want to have large amounts of incorrect responses, which will serve to waste time and limit the value received. In order to achieve both, the model should have high “precision” and “recall” scores.

Precision is how often a model incorrectly identifies information (i.e. mislabeling a waiver of subrogation clause as a waiver of consequential damages). Recall is whether the model identifies all instances of an item (i.e. missing a potential cause of delay such as rain from daily reports). Good models will have high precision, as well as high recall scores. Do not be afraid to ask for any platform’s scores for each of its algorithms.

It is also important to consider the ease of interaction for users and integration with current systems. An accurate and sophisticated platform that is difficult to use and not integrated with the current systems likely will not be used by the team or will be incorrectly used.

Last but certainly not least, is security and privacy. Make sure that the platform where information (whether in the AI/machine learning space or otherwise) is deploying best-in class cyber security measures/practices and will keep company information private. Be sure to have information technology partners and vendors review any proposed platform to confirm that it complies with internal requirements and any commitments for privacy and security that may have made to third parties such as owners or other end users.

Once the right platform has been identified and implemented, projects can be assessed more intelligently, yielding long term results of a more efficient and safer job site. As AI and machine learning continue to advance, it’s only a matter of time before more and more construction companies start to reap the benefits of these powerful tools.