Embracing the Future: How Three AI Solutions Are Shaping the Workforce and Asset Management 

We are only just beginning to experience the full potential of artificial intelligence. As we navigate the evolving landscape of artificial intelligence, it’s clear that AI is redefining how we work. In asset management, AI promises groundbreaking advancements that enable humans to do more of what they’re good at while leveraging machines that have their own capabilities. 

Electricity took several decades to settle into its place as user adoption increased over time. These decades are the “In Between Times,” according to  Ajay Agrawal, Joshua Gans and Avi Goldfarb in their book Power and Prediction: The Disruptive Economics of Artificial Intelligence. AI will have an adoption cycle, too. Whether it lasts for decades, like electricity, is for the AI experts to estimate. But for the rest of us, we can expect the market to offer three types of solutions, including the point solution, the application solution and the system solution. Knowing the difference between these could make a huge difference on which AI-enabled software you decide to learn, try or buy. 

 Three Types of AI-Enabled Solutions 

As AI use grows, there is a division of labor that needs to occur in several domains, including asset management. Humans have a better oversight for judgment, abstract thinking and strategic decision-making. We can delegate a substantial amount of routine cognitive work to machines, freeing up more time for us to focus on what we excel at. Machines have proven their ability to process more information than we can, faster and unemotionally. As such, they are better able to do what humans aren’t particularly good at – make predictions. In Power and Prediction: The Disruptive Economics of Artificial Intelligence, Agrawal, a renowned economist and professor at the University of Toronto, known for his expertise in artificial intelligence and its impact on business and economics, and his fellow authors describe three types of artificial intelligence solutions: 

1. The Point Solution 

The point solution involves the simple replacement of an older prediction technology with new AI-enabled tools. A good example of this is vibration analysis. This practice has been around since the 1960s. During the early days, a technician would walkthrough a facility and take readings from each collection point. Transferring the collection from the datalogger to a desktop computer, the technician would evaluate the data, conduct the analysis and make recommendations. Over time, we’ve been able to increasingly embed some of that human knowledge into analysis. The human focused more on judging the results, making recommendations and decisions. Today, AI has been useful for improving pattern recognition, signal processing improvements and automating fault classification, just to name a few. This is an excellent example of how AI has taken over parts of the process and brought significant improvements and efficiencies to a long-standing approach to predictive maintenance. 

2. The Application Solution 

The second type of AI solution is the application solution. These are the types of applications that enhance existing products or processes. A good example of this is Natural Language Processing.  

A paper titled ‘Where Do We Start? Guidance for Technology Implementation in Maintenance Management for Manufacturing,’ presented at the ASME 2019 International Manufacturing Science and Engineering Conference stated, “One of the most exciting recent developments is in natural language processing to enable work order texts to be read and analyzed more efficiently by computers.” This improves an organization’s ability to harvest valuable information from the free text field of a work order filled in by a skilled craftsperson.  

The implications of applying these capabilities can include technicians spending less time in the application, selecting data from dropdown menus. Instead, they can simply type (or even speak) their observations and conclusions in free text instead. These NLP-enabled applications will do the work of extracting the necessary data points, saving significant hours of expert labor. 

 3. The System Solution

 The system solution is pervasive. In the same way how electricity was invented, used by early adopters and eventually built into a network that delivers electrical energy to homes and businesses, computational resources have found their place in the market and have become ubiquitous. With the same effort that goes into flicking a light switch, we speak into our devices to give us directions to the nearest sandwich shop.  This is the essence of what Jeff Bezos described about AI at the 2024 New York Times DealBook Summit. He said, “There isn't a single application that you can think of that is not going to be made better by AI.”  He supports this claim by reminding us how electricity enabled industrial development at scale, and how cloud resources have enabled digital applications to scale. Similarly, the applicability of AI will be available at scale. We’re collectively begun to experience this today as user adoption for applications like ChatGPT increases every day. Soon there won’t be an application that isn’t plugged into the AI layer in a similar way that our devices are plugged into the electricity grid.  

Know What You’re Signing Up For 

In the asset management domain, big promises are being made in the market. Some solution vendors claim the ability to automate root cause analysis. Others promise their AI understands complex relationships and can make some bold predictions. Some describe the comprehensive capability of machine learning and physics-based simulation. While some of these might one day be possible, it’s important to consider the dependencies that come with them. To get started, you’ll need massive amounts of data from multiple data sources, and a data operations pipeline that manages: 

1. 1.  Data modeling applications and expertise 
   2. The data quality assessment process
   3. Prognostics modeling applications and relevant data science resources  

Using the example of automating root cause analysis, we can break down the key dependencies that need to be considered. First, that the necessary data needs to be captured with sufficient quantity and quality, including equipment and process specifications and historical failure records with timestamps. Next is the qualitative data collected from people providing their own observations, process flow documents and procedures. Then, we’d need to embed the fault-tree approach - the ability to distinguish between symptoms and actual root cause, and the process of generating recommendations, assigning them to people and to a timeline.  

There are other implications that are far too numerous to fully discuss in this blog post, but most readers will begin to understand the complexity of even coming close to handing over root cause analysis to an AI-enabled application. 

The key recommendation is that asset management professionals should be able to use this knowledge about the three solution types to decide how the AI-enabled technology they’re considering can be applied to their own asset management system. Is it a simple replacement of what already exists today, or does the application come with a promise that has system-level implications? 

Use AI Today to Protect Assets 

One thing is certain about AI and its usability: it will require significant ‘hand-holding’ from humans before it can do some of the things described in the examples above. For example, large language models perform well when tested against a vetted, generalized body of knowledge in the topic of reliability and maintenance. However, if we expect computers to be observant and make specific maintenance recommendations, we will need to embed the models with more human knowledge to make them reliable.   

AI is available to you today in the HxGN APM application. It is more than a point solution because it doesn’t replace something that currently exists. It is an application solution because it produces something that doesn’t exist – visibility to cost, risk and performance. HxGN APM is built with a pattern protections capability that provides your organization with machine learning to take an important step toward optimizing the division of labor. HxGN visual AI is purpose-built to begin embedding the knowledge of your organizational experts into the process of evidence gathering and decision-making. We’d enjoy the opportunity to walk you through a use case and let you put your hands on it. Please read more about it on our HxGN APM website.