Request A Demo
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Deep learning can process thousands of combinations quickly and efficiently. It can identify patterns and relationships that humans may not have previously considered (e.g. Site engineers with less than five years of experience are more likely to contribute to cost overruns due to a higher number of quality defects). In general, the more data available, the better the predictions within a project.
However, there are important limitations to consider.
You might think that an accountant in the head office has nothing to do with quality issues in tile work on site. Construction projects are typically organized into departments, and we’ve long assumed that these teams operate independently. In reality, when people are brought together to achieve a shared goal (in our case it is delivering the project on time and within budget), everyone affects one another. The difference lies in the size of the ripple effect. Some roles influence certain activities far more than others.
Consider an accountant who consistently delays subcontractor payments. As a result, materials arrive late on site. To recover lost time, the site team rushes execution, ultimately compromising quality. Collecting, cleaning, and mapping data takes time and effort. Therefore, it is essential to be mindful of the cost per unit of analysis and ensure that the benefits of improved prediction outweigh the cost of collecting and managing the data.
Another key factor is the law of diminishing returns. While more data leads to better predictions, the improvement does not increase at the same pace.
For example:
Although more data improves prediction models, the rate of improvement is not linear.
Machines can outperform humans when sufficient data is available. A well-known example is loan approvals, where AI models have predicted borrower default with higher accuracy than humans. Human judgment can be influenced by emotions, mood, and cognitive bias, whereas machines remain neutral. Some job requirements are largely objective (for example, sales roles with defined targets). In construction, people are the greatest asset on any project, and their performance is often relational rather than solely objective.
While the site team is accountable for meeting the schedule,delays are inevitable. What often determines success is not technical capability alone, but communication, leadership, and responsibility. I have personally worked with individuals who had strong experience and technical skills but they were extremely difficult to work with. This friction affected overall performance. This is why many companies are increasingly shifting their focus toward relational requirements, not just technical ones. An AI model will not assign “communication scores” to team members but these qualities influence progress and outcomes.
For this reason, it is not advisable to apply an AI model developed for one company directly to another. Organizational culture, relationships, and ways of working differ, and those differences affect performance. As AI models gain access to more data and process more combinations, human predictive ability diminishes in comparison.
The first step in adopting AI is digitaliz organizational operations and moving data to the cloud. However, this is still not the reality for most contractors. Many organizations rely heavily on on-premise ExcelSheets and folders stored on local hard drives. They also depend on Emails, WhatsApp messages, and paperwork for communication and approvals.
Once data is moved into cloud-based applications, the next step is data architecture. Data must be mapped, structured, and prepared for AI use. This process requires time and effort to be done correctly. However, many organizations feel impatient. They (like most of us) operate in a constant state of urgency and immediate gratification, rushing through project chaos to achieve the goal of completing projects on time and within budget. Over time, as projects become more complex, the pain of managing them becomes louder and more evident.
Managing and organizing big data should be seen as a long-term foundation. Months are spent building a strong foundation to support a tall and durable building, or as a tree develops deep roots to grow stronger and live longer. Similarly, data foundations are essential yet largely invisible. Structuring project data before building any application or AI model does not immediately appear in the front end. Because it is not visible, it often feels unnecessary or like a waste of time but it is the foundation for future success.
Ultimately, construction management is about managing risk. Construction projects typically face three types of risks:
These are risks that are easy to identify, such as when an activity exceeds its budget. Both humans and machines can detect cost overruns when cost baseline and actual cost data are available. However, machines can perform this analysis faster, especially when detailed drill downs are available (like in Power BI).
Humans can recognize patterns intuitively. An experienced engineer may sense that something is wrong in a report before identifying the exact issue. This “gut feeling” results from years of accumulated experience and cannot be easily programmed into a machine. Humans have the upper hand in dealing with these risks.
These include rare and unforeseen events such as force majeure. Both humans and machines largely fail to predict this category of risk.
To achieve the best results, humans and machines must work together. Machines should handle large-scale data processing and prediction, while human attention should be applied where intuition, judgment, and experience are required.
Today, there is an abundance of tools available from companies such as Microsoft and Google that enable the use of AI in construction. While these tools offer significant potential, they also come with limitations and risks.
By understanding these constraints and designing systems thoughtfully, organizations can unlock the full potential of AI and generate more accurate, reliable predictions over time.
