Let’s acknowledge the reality: Capital expenditure (CAPEX) projects rarely ever come in on time and on budget. Project schedule and cost overruns are the norm, not the exception. But this is changing with artificial intelligence. It’s providing a project risk analysis with an entirely new layer of project certainty.

From an owner’s perspective, risk analysis means trying to determine that the amount of capital expenditure allocated to a venture is appropriate. 

For example, the most cost-sensitive variable in our project is time. We also have to take into consideration that CAPEX projects have a well-earned reputation for running over with regard to time. 

Often, we have competing work and phases with limited resources. There’s also a huge degree of dependency between the sequence of design, procurement, fabrication, delivery, installation, commission and closeout. Such a highly sequential set of dependencies really puts owner organisations at extreme schedule risk. 

But what if we could predict the risk and uncertainty of any given project upfront and set our expectations closer to reality right from the beginning? Driven by today’s construction scheduling software using artificial intelligence (AI), project risk analysis is delivering an entirely new layer of project certainty to owners. 

The ‘best-case scenario’ trap

Many are familiar with the traditional risk analysis process where we start off with a deterministic forecast. We strip out any contingency that’s already built into that forecast because, ultimately, the percentage of contingency is what we’re trying to ascertain. 

We then model the risk through two variables: uncertainty and actual risks. A risk analysis simulates the execution of the project thousands and thousands of times and then produces the risk results. 

Unfortunately, traditional estimating and scheduling tends to generate a best-case scenario rather than a most-likely scenario

This is because a planner or scheduler will develop a schedule forecast, but very rarely will that forecast include some of the potential unknowns in the form of risk events that could potentially happen during execution. So, the plan that is created assumes uneventful, perfectly executed events. And as we all know, best-case scenarios very rarely happen. 

How AI can help

One huge step forward regarding AI involves eliminating some of the statistical complexity that legacy risk analysis carries with it. 

Previously, we would interview team members and ask them for three-point estimates, or three-point distributions of best-case, most-likely and worst-case scenarios. Well, that’s actually very difficult to extract out of a team member. 

But with AI, we have large quantities of historical data at our fingertips and we are no longer just limited to three-point estimates. For example, if we have ten historical projects in our AI knowledge library, we can end up with an input distribution that has ten points, not three points. So, it becomes much more accurate. 

Another challenge with traditional risk analysis has been to accurately capture those inputs. With AI, we have the luxury of leveraging historical project information throughout the project lifecycle. 

Computers can now review prior project schedules, cost estimates and even previously generated risk registers, and start making suggestions back to the planner and the scheduler during the planning process. This eliminates the need to build a plan and then reactively come back and throw contingency at it later.

Overall, with AI, we’ve reduced the overhead and the complexity of the development of a risk register by allowing the computer to make suggestions based on historical risk events that have occurred before that impact schedule and cost. 

That’s an incredible step forward because now we can start to build our risk register on human expertise and on the suggestions of AI, calibrating a risk register based on the best of machine and human input. 

This partner content was brought to you by InEight. To learn more about InEight’s risk analysis and construction scheduling software, click here.

1 Comment
  1. Michael Wallace 1 month ago

    Yes, removing human bias would probably increase contingency (it’s a good thing). However, I believe there are greater benefits in AI when looking at company culture and human behaviours. AI would be great for monitoring performance when implementing management of change (MoC) within the organisation and determining the cost-benefits. The feedback on MoCs can then be used in revising KPIs and improving business profitability.

Leave a reply

Your email address will not be published. Required fields are marked *

*

©2020 Infrastructuremagazine. All rights reserved

CONTACT US

We're not around right now. But you can send us an email and we'll get back to you, asap.

Sending

Log in with your credentials

Forgot your details?