Using Artificial Intelligence in Risk Management

One of the challenges that the construction industry has is a lack of coordinated data sharing.  Some agencies have started to collect large amounts of data, but it tends to be project specific, not particularly accessible and not organized in a way that it can be easily analyzed to develop usable information. 

Data mining could be particularly valuable for risk management and improving the quality of design and planning contracts.  If an agency had multiple years of change order data from its construction contracts it may be possible through data mining to extract information as to the types and severity of changes.  This information could be studied to identify reoccurring problem areas.  For example, if it was determined that large and consistent geotechnical related change orders occurred, it might point toward changing the design and planning process to require more geotechnical investigation. 

This same data mining could help categorize changes so that the severity and probability of various types of changes could be better understood, allow for better contingency forecasting, and promote better planning for how to handle individual changes.  For example, knowing that weather related changes occur on a certain percentage of a certain type of project and have a certain average impact would provide a better foundation for predicting the impact of these changes on individual projects and encourage more effective contract language and decision making for dealing with these impacts.

If such data is available, and before analysis starts, it needs to be reviewed to see if individual change descriptions accurately explain the reason for the change (for example, is there a bias against stating that there was a “design error” or the owner “changed criteria”) and whether the data is sufficiently detailed to extract usable keywords and phrases.  This exercise would also help contract administrators better coordinate with management to identify key data that should be tracked in the future and promote consistent change order language.