Joseph Berti, VP Offering Management, IBM AI Applications looks at how IBM is using AI to manage critical infrastructure.

As the COVID-19 pandemic continues and the specter of recurrent closures of businesses, campuses and facilities remains, at IBM we are always vigilant about reality that many of our clients cannot shut down or have their employees work from home.

The reason is that many of our clients manage critical infrastructure, from government buildings needed to ensure society can still function to utilities that ensure our power stays on. These organizations need people onsite to ensure that vital equipment is working or to make critical repairs. They must also remain compliant with social distancing and occupancy guidelines in facilities that may not necessarily have been designed with these needs in mind.

Unfortunately, organizations that manage this critical infrastructure – and even many of those that do not – cannot afford to simply ride things out and return to business as usual. Factors like adverse weather, population growth and the prevalence of international travel all suggest that outbreaks like COVID-19 will become more common, not less. An outbreak can now move from a remote village to a large city on the other side of the world in approximately 36 hours, some research suggests.

The three “layers” of AI

When we think about how organizations should be applying AI to solve these problems, we view AI as operating at three distinct layers.

The first layer is using AI to monitor facilities, and this is already pretty easy. AI models trained in anomaly detection are common, effective and can be applied at scale to alert engineers any time a piece of equipment isn’t working the way it normally does. In this way AI can also facilitate more remote monitoring, lessening the need to bring everyone into a given facility on a given day. However, whether a piece of equipment is working or not only tells part of the story. AI can also allow us to go deeper to reach a second layer of analysis.

AI, for example, can put an anomaly into a broader context by letting us holistically assess an asset’s health. We can keep track of the asset over its entire lifecycle, from installation to its retirement, using AI to create a determination about the asset’s health at a given moment. Is that equipment acting up because it’s due for inspection or repair anyway, or might there be a more serious issue?

Finally, AI has also proven that it can help organizations achieve a third layer of analysis, predictive forecasting. It can analyze how often a given piece of equipment typically breaks down or how it might deteriorate over time, taking into account factors like how often the equipment is used and the environment where it is stored. By looking at an asset, its history, current health and the performance of its peers, we can make predictions about what repairs might be needed not only in the next few months, but the next several years.

The power of predictive forecasting

There are several reasons why this level of analysis will be necessary for large organizations going forward. First is that the needs vary widely. Many workers happily work from home and roughly half say they are even perfectly happy to forgo going into the office indefinitely. However, there are many industries, from food processing to nuclear power where this is obviously unfeasible. Keeping such facilities open with fewer employees will require being able to be more strategic about where those employees work and how they spend their time.

Predictive forecasting can also reduce costs and promote sustainability. The longer a piece of equipment is used in sub-optimal conditions, the more expensive it will be to repair and the sooner it will need to be replaced. Predictive maintenance can help ensure issues are resolved before they become serious, reducing downtime and ultimately promoting longevity. The longer we can make these devices last, the less energy and the fewer natural resources we will consume replacing them.

AI has the capacity to augment human thought and creativity, abilities that will continue to become all the more valuable. But in thinking specifically about the problem of managing physical assets, it’s important to distinguish between the many different layers of analysis that AI is capable. By using AI to not only monitor, but to asses and predict as well, we can reduce the toll of these closures on essential workers while ensuring workplaces are better-prepared to weather the challenges to come.