Five Ways Industrial AI is Shaking Up Manufacturing (and Who’s Doing It)

In my previous blog, I took the liberty of extending Marc Andreessen's Software is Eating the World statement to how Artificial Intelligence was now eating software, implying that AI was becoming increasingly more pervasive in today's software businesses. I also described industrial AI as emerging not just as a technology trend or a buzzword but also as a crucial piece of the puzzle in transforming manufacturing.

This blog takes a step further and examines industrial AI in its various forms, from machine learning and deep learning to Copilots and Causality to vision systems and robotics. Specifically, I'll explore some of the prevalent use cases for industrial AI and list some software technology vendors providing these capabilities.

1. Asset Health and Maintenance 🛠

Over the past few years, Asset Performance Management (APM) has always yielded some of the most common low-hanging fruit use cases for analytics and, by extension, Industrial Transformation. Given the rise of inexpensive IoT sensors, collecting vibration, temperature, and pressure data from assets has gotten substantially simpler. As a result, anomaly detection, unplanned downtime, and asset reliability are some of the commonly chosen use cases to prove some quick wins and scale their use cases.

Consequently, it's no surprise that most software companies assessed in our upcoming Advanced Industrial Analytics (AIA) Solution Selection Matrix (SSM) provide some level of AI-based asset monitoring and analytics capabilities, with a few extending it to energy monitoring. To avoid listing almost every vendor in the market, I will refrain from calling out specific providers here.

However, a sub-category of vendors focusing exclusively on AI-based machine monitoring is worth mentioning. Arundo, AssetWatch, Augury (now expanded to Production Health since acquiring Seebo), Dynamox, Infinite Uptime, KCF Technologies, Nanoprecise, Petasense, Samotics, Sensoteq, Tractian, Viking Analytics, and Waites are some of the key players in this niche that provide AI-powered machine monitoring through a combination of software and hardware.

2. Process Analytics 📈

Process Analytics is another strong candidate for Artificial Intelligence-led disruption in manufacturing. However, it's a little less prevalent than APM since mapping process parameters tend to be more complex than collecting vibration data from IoT sensors. The scope of AI in process analytics mainly involves regression, classification, and even clustering models to uncover patterns and trends in material flow, batch recipes, quality parameters, etc.

A variety of industrial technology providers serve this space of AI for process analytics. On one side, companies like Seeq and TrendMiner, which have emerged as the go-to tools for Process Engineers (replacing Excel), provide machine learning-based analytics on time-series data to analyze trends. Additionally, companies like Oden provide predictive and prescriptive insights for process engineering and execution, targeting process engineers and frontline operators.

On the other hand, companies such as Augury, C3.ai, Falkonry, SparkCognition, SymphonyAI, and TwinThread leverage machine learning (and, in some cases, deep learning) models to analyze process data in context with asset data, thereby applying industrial AI to uncover more complex relationships across maintenance, reliability, quality, planning & scheduling, energy optimization, batch release, etc.

3. Process Control ⚙

While several companies leverage AI for monitoring and analyzing process parameters, only some extend their capabilities to impact process control by writing back to HMI/SCADA and other execution systems. And understandably so, since this requires a much deeper level of data integration and comes with a higher risk profile.

For instance, a false positive prediction might not significantly threaten operations. However, automatically changing process set points and batch recipes based on that prediction could potentially result in suboptimal batches, production stoppages, and unplanned downtime.

Applying AI for process control is a complex but necessary step towards Autonomous Operations, an increasingly common goal for many manufacturers today. Canvass AI, Oden Technologies, Quartic.ai, and TwinThread are some of the industrial AI companies that have stepped up to this challenge, providing machine learning and causal AI models that give not just process monitoring and analytics but also prescribe recommended set points to impact process control.

4. Productivity ⏰

Applying Artificial Intelligence to improve productivity is not a recent phenomenon; automating repetitive and menial tasks has always been one of AI's most significant value propositions. However, a coincidental occurrence of two singular events in the recent past has enabled AI to play a much more substantial role in improving productivity.

To begin with, the spectacular rise of ChatGPT and other large language models and their ability to consume, interpret, analyze, and generate large volumes of text grabbed significant adoption and venture capital attention. Additionally, the rise of these foundation models coincided with the post-COVID Great Resignation of 2020-21, during which millions left the workforce.

This skilled labor shortage, combined with an increased attrition rate, resulted in a dramatic shift in demographics in the manufacturing workforce. According to LNS Research's recent study on the Future of Industrial Work, the average tenure of a frontline worker in manufacturing went down from 20 years to just over three years between 2019 and 2023. As manufacturing companies gradually realized they could not return to pre-COVID employment levels or get back their 30-year veterans, they increasingly turned to Artificial Intelligence to fill the knowledge gap.

Since then, most industrial technology providers have partnered with big tech to leverage large (and small) language models to embed generative AI capabilities into their product portfolio. Over time, techniques like retrieval augmentation generation (RAG) and a focus on traceability have been applied to address common challenges, such as hallucination and data security.  

As of this writing, most applications of generative AI models in manufacturing center around information retrieval and summarization, data visualization, programming, translation, root cause analysis, etc. Notably, C3 is leading the charge here, focusing on Enterprise Search and an increasing list of industry-focused generative AI-based applications.

Almost every single vendor in the industrial technology landscape has released its versions of Copilots and AI Assistants over the past couple of years to assist users with information retrieval, summarization, and brainstorming. More recently, the focus has shifted towards agentic AI— leveraging AI not just as a tool for automating repetitive tasks but as a self-learning system capable of understanding and streamlining complex workflows and driving autonomous decision-making. While a few vendors have started to offer these agentic AI capabilities, there are still ways to go for the industry to have access to truly autonomous agents with these complex decision-making capabilities.

5. Worker Safety 🦺

Finally, industrial AI is also being applied to frontline worker safety in factories across the globe. As of today, a small number of companies leverage computer vision in combination with neural networks and natural language processing to improve worker safety while also focusing on work instructions, ergonomics, productivity, and process quality on the factory floor.

Emerging companies like Intenseye, Leela.ai, Retrocausal, SparkCognition, Symphony.ai, and Tulip provide some of these hardware and software offerings, which use vision-based Artificial Intelligence models to monitor and optimize workforce productivity, safety, and performance.

On the other hand, robotics companies like FANUC, Fetch Robotics, OTTO robots, Universal Robotics, etc., combine machine learning with 3D cameras and LiDAR sensors to capture information, analyze it, and make inferences based on their environment and mission. These robots, cobots, and AMR/AGVs go beyond just safety and can also empower users to work alongside these systems to manage complex tasks that cannot be entirely left to an AI system (yet).

Table 1: Industrial AI Vendor Landscape

Summary and Recommendations

Manufacturing is one of the most data-intensive industries, and artificial intelligence is poised to disrupt it. As of this writing, early-stage overhype has given way to more realistic expectations as the industry grapples with data quality, security, time to value, and scalability challenges.

While most industrial AI providers have played the hype cycle well and capitalized on artificial intelligence's exponential growth, a few have chosen to stay more pragmatic and not play into the hype. Nevertheless, LNS Research believes that the true industrial AI winners will be the ones who provide practical solutions that integrate with existing systems and solve actual business problems, such as the ones mentioned above, in a scalable way.

In addition, emerging branches of Artificial Intelligence are finding their way into manufacturing. Causal AI, which has its roots in the fin-tech world, is increasingly used to find underlying causes (as opposed to correlations and patterns) to uncover how changes in one variable will influence others on the factory floor. Another emerging category that LNS Research believes many industrial AI initiatives will be well-served by focusing on is leveraging knowledge management and agents to drive Intelligent, AI-driven decision-making across multiple layers in the organization.

Over the next few years, industrial AI will continue to play a critical role in defining the leaders, fast followers, and laggards in the next generation of manufacturing. Companies that embrace and welcome artificial intelligence (with recommended prudence) will likely have substantial potential for operational and financial gains. In contrast, the others will increasingly see themselves falling behind and potentially ceding manufacturing as a competitive advantage.

Here are some recommendations for manufacturers to consider as they enter the industrial AI race.

• • • Lead with use case and user persona impact: Most strategic initiatives fail to scale because they are rooted in technology and do not solve business problems. Similarly, while industrial AI can potentially disrupt several aspects of manufacturing, it is critical to approach it by considering C-suite objectives, existing strategic initiatives, and use cases. However, readers should be aware that focusing too much on use cases without considering the impact on key user personas can lead to sub-optimization.

    • Good AI requires good data:  As many manufacturers continue to fast-track AI initiatives, it is important to note that a robust Operational Architecture, built on a solid data model with high-quality and contextualized data, managed and governed by Industrial DataOps, is essential to getting the promised results from industrial AI.

    • Industrial AI is here to stay, but so are the people: The manufacturing industry is not the first, nor will it be the last, to hear the AI-is-going-to-take-away-jobs industrial AI's ability to automate repetitive tasks and enhance human-machine-software interaction will most likely impact the manufacturing workforce to some extent.

However, more importantly, industrial AI will improve the nature of most jobs and create various new ones, exacerbating the current skilled worker shortage. To meet this increased demand for skilled labor, manufacturing companies will need to empower their workforce to upskill and cross-train themselves in the near future.

In other words, people will not lose their jobs to AI, but they will probably lose them to the next person who can work with AI.

*All vendors mentioned are listed in alphabetical order.