For the past 50 years, large capital projects have relied on the science of what is known as CPM (critical path methodology) to plan and forecast project completion dates. There have been iterative improvements to CPM, such as accounting for risk in the form of generating risk-adjusted schedules, but the fundamental approach to CPM hasn’t materially changed. So, has CPM really proved itself as a worthy planning approach, or is a refresh needed? Let’s face it, capital projects don’t exactly have a great track record for on-time completion—perhaps there is room for improvement?
The Fundamental Problem with CPM Isn’t Actually CPM
The philosophy behind CPM is sound: break down the scope of work into activities, estimate their durations, and then let CPM calculate the resultant dates based on their relative sequence—easy. Easy, yes, and mathematically sound, but only if the duration and sequence inputs are correct. In reality, that is a huge IF, as the entire calculation is 100% reliant on the inputs being correct. Those inputs don’t come from a computer that operates in a binary yes/no-type world; those inputs come from a human expert typically known as a planner.
The planner carries two responsibilities: they have to understand how to drive a CPM software tool and they have to understand the duration and sequence inputs for the project they are planning. The first is easy—that’s just training in a software tool. The second, though, requires domain expertise in the thing that is actually being built—the project itself!
That domain expertise is incredibly hard to capture and typically takes years to establish. In most cases, it involves the planner actually having hands-on, in-person experience of working on-site during construction execution. This expertise and knowledge isn’t something you can gain sitting behind a desk in an office.
Interaction Between Humans and Computers Is Backwards
Traditional use of a computer has been dependent on the human (a planner) giving the computer (the CPM software) instructions and input (durations and sequences of work). From there, the computer then does its calculations and returns the answer in the form of a project schedule.
The computer cannot do any of this without being told what to do by the human. The computer is guided by the human and, in reality, is a consumer of information from the planner and not a provider of knowledge. Consider for a moment if we reversed this, though, and the computer was able to actually provide guidance back to the planner? What if the computer could offer suggestions based on, say, past project performance? This is exactly where the rapidly emerging science of AI (artificial intelligence) is making a big impact in the world of project planning.
An Artificial Intelligence Primer
There are many definitions of AI. All revolve around the concept of a computer having the “ability to perform tasks that normally require human intelligence.” At its most simplistic, an AI algorithm is based on learning from the past.
This fits in very nicely with the science of project planning. If a computer can capture historical project performance and subsequently offer up suggestions from this during the planning process, then the concept of the human driving the computer gets turned on its head. For the first time, the computer can now offer suggestions to the planner and can assist in decision-making through intelligent guidance and suggestions.
The key to AI being effective during the planning process, though, is the ability for the computer to understand context—context of project type, location, scope of work, crews being used, and so forth. This is where neural networks and inference engines come into their own. By not being solely reliant on such things as activity description matches, an AI inference engine can make informed comparisons of similar past work purely based on analogous context.
AI and Project Planning
The past 18 months have seen the launch of several AI-driven planning CPM solutions. These tools not only provide a means of easily capturing historical project data (schedules, risk registers, and standard productivity rates) but more importantly, through AI, provide realtime guidance from the computer to the planner during the plan creation process. These tools literally can now use historical information to help better predict the future.
The tools that will ultimately win this race for becoming the standard in next-generation CPM will be those that can actually get smarter in their suggestion offering. Through what is known as “machine learning,” these AI planning tools will need to be humble enough that they are capable of being recalibrated based on the expert planner. By understanding whether or not the suggestions that it is making are correct (and then adjusting accordingly), the computer will become more and more valuable to the project planning community.
While CPM planning tools have had several facelifts over the years, the basic premise hasn’t changed—that is, up until the recent arrival of mainstream AI. With AI, planning tools are now starting to provide expert guidance to the planner during the plan creation process. This is a monumental leap forward for CPM. This doesn’t in any way replace the expertise of, or take away plan ownership from, the planner; instead, it augments and assists in their ability to build more realistic project forecasts. That is an undeniably positive step forward in the science of project planning.
To learn more about AI-based planning and scheduling solutions such as InEight Basis, or to schedule a free demo, visit InEight.com.
Dr. Dan Patterson is chief design officer with InEight. In this role, he focuses on expanding upon his vision of creating next-generation planning and scheduling software solutions for the construction industry. Dan is a certified PMP (Project Management Professional) by the PMI (Project Management Institute).