Solving Project Challenges With AI and Machine Learning
Reposted with permission from constructionexec.com, June 5, 2019, all rights reserved. Copyright 2019.
Whether it be a safety violation, work that is done in an unexpected way, materials turning up at the wrong time or in the wrong place, or work not being completed on schedule – in construction project management, atypical is almost always bad. Project managers strive to make construction delivery as smooth and predictable possible; errors and exceptions are costly and increase the risk of a project being subject to a cost or schedule blowout.
Today, AI and machine learning methods can identify construction errors and handle exceptional site conditions fast. A project manager can get a text with information about a crew working without safety gear in less time than it takes to walk the perimeter of the site. In this way, project managers know precisely what matters on site and when. Critical issues can be prioritized and addressed in an efficient, rapid-response style. AI can give a project manager more eyes on site to monitor more critical work in the same amount of time.
For those on the fence about embracing AI and machine learning, there are some useful considerations when identifying those areas in a business most ripe for an AI or machine learning pilot.
FINDING THE RIGHT FIT FOR AI
Taking the time upfront to correctly identify which project challenges AI and machine learning may be able to fix is critical. Selecting the wrong AI pilot can be doubly bad. First, it can accrue financial losses: money gets spent on a pilot and the returns are not up to par. This is never desirable. Second, if it’s a company’s first foray into advanced tech, a failed pilot can have an outsized cultural impact. It is very easy to sour on a new idea when it doesn’t live up to its hype.
- Where is money or time or effort being hemorrhaged? It’s always best to start with the most painful of pain points and work backwards from there.
- Where are project estimates and projections performing best? Where are they way off? Estimates that are consistently far from actuals often mean that an organization is relying on old assumptions about key performance drivers. Machine learning is particularly good at making accurate predictions based on complex patterns. Leveraging machine learning for project estimates is a classic use case.
- What in not known about performance that should be? If there’s something that should be measured but isn’t, it’s likely important. If it can be measured, it can be analyzed. If it can be analyzed, it can be improved.
- What’s the dollar figure in either cost reductions, increased productivity or revenue that makes a pilot worthwhile?The answer to this question will help determine whether a given machine learning solution will generate the return that’s needed for full organizational buy-in.
What follows is two brief examples demonstrating how AI and machine learning can help solve challenges in project execution.
ENHANCING QUALITY AND SAFETY ON THE JOB
Maintaining project quality and site safety are fundamental for effective project delivery. The latest computer vision technology helps the QHSE team to stay one step ahead. AI-enabled site monitoring allows site inspections, quality assurance, and safety monitoring to be faster and much more thorough. A series of IoT-enabled cameras trained on critical areas of the site, for example, offers two main benefits:
- the QHSE team gets a real-time data feed of key site areas; and
- machine learning algorithms are continually running the background to assist with the detection of quality and safety violations.
When a potential violation is identified, an automated alert is sent out to all relevant stakeholders. In this way, construction can be closely monitored without deploying an army of inspectors on site.
When used to its full potential, AI allows for fully automated and remote site inspections and enables safety risks to be identified before they turn into full-fledged safety issues. Similarly, with AI, as soon as something is not built to spec, an alert can be triggered. Documenting and mitigating quality or safety violations will no longer require manual inspection.
IMPROVING ESTIMATION AND PROGRESS MEASURES
A huge focus of any project controls team is the creation, and rolling revision of, cost and schedule estimates. These estimates are used to build constructions schedules and capital plans.
Making accurate project estimates is often a huge challenge. Machine learning is great at estimating and predicting future behavior, it can recognize hugely complex patterns in the execution of prior projects and apply this knowledge to future projects. Better estimates mean better plans. Better plans reduce the likelihood of a project missing milestone dates or being liable for penalties or expensive liquidated damages.
Another exciting area in machine learning right now is real-time progress measurement. The technology here is new and the potential is enormous. AI-enabled progress measurement involves setting up internet-enabled cameras and LIDAR devices on site to process real-time site footage with reference to an existing 3D BIM model. The devices continually compare activity on site with a digital twin and trigger alerts when discrepancies are found. Instead of manual quality inspections and long checklists, sites can be monitored remotely and automatically, allowing for construction errors to be identified in near-real time. Construction errors or omissions can be spotted immediately.
The field of AI is booming right now. It can be tough to know exactly how and where to begin. A good first step is convening an AI team to advise on advance tech opportunities. This team should recruit from a wide range of disciplines. Individuals with advanced tech knowledge should be part of the team, certainly, but just as importantly, the voices of ‘boots on the ground’ project managers must play a prominent role. The best laid plans fall by the wayside if they don’t make life on site convenient. Project managers know their work intimately and will know which solutions are likely to be the most useful.
On the other end of the spectrum, younger and less senior team members are often more closely exposed to and familiar with both the latest technological thinking as well as precisely how and why it may be relevant. They should have a seat at the table when discussing AI initiatives. It’s always good to let the ideas trickle up.