The AI Co-Pilots Are Coming. What Are They?

March 12, 2024
Not to be confused with their digital assistant predecessors, artificial intelligence co-pilot tech leverages multi-disciplinary datasets for analysis and insights to help manufacturing workers and potentially complete large portions of work independently.

Amid all the attention over artificial intelligence (AI) over the past year and half since the wide release of ChatGPT, there’s been a co-development of what are often called co-pilot technologies. The co-pilots being developed for industry are akin to the digital assistants we’ve been hearing about for years now that have had some limited success but not yet been widely adopted. 

The key differentiator between these new co-pilots and digital assistants is the quickly advancing development and use of large language model (LLM) AI technologies, of which ChatGPT is the most well-known. This has the potential to impact automation professionals in that the co-pilots being developed for industrial applications are aimed at improving human-machine collaboration in the manufacturing and processing industries.

To say that we’re still in the early days of LLM AI use in general, and co-pilot technologies more specifically, is an understatement. But co-pilot products are on the way and you’ll be hearing a lot more about them over the next few years. Considering how quickly AI tech is already shaping corporate plans and some aspects of the workforce, keeping up with developments around these technologies makes good sense given their potential to heavily influence how people interact with automation technologies of all kinds.  

Artem Kroupenev, vice president of strategy at Augury, a supplier of AI technologies to predict and prevent machine failures as well as optimize production quality and throughput, explained that AI co-pilots are “intelligent assistants that empower workers to plan and complete tasks efficiently and drive better decision-making by leveraging vast, multi-disciplinary datasets for analysis and insights. With time, AI co-pilots will be able to effectively collaborate [with humans] at any level of task planning and execution and complete large portions of work independently.”

What sets this AI tech apart?

Beyond their use of advancing AI technologies, another difference between co-pilots and the digital assistant technologies we’ve been hearing about for years lies in the concept of automation vs. autonomy. 

“Unlike their [digital assistant] predecessors, which focused on automating simple tasks, today's AI co-pilots are equipped to tackle complex problem-solving, decision-making and creative processes,” said Kroupenev. “This shift marks a pivotal evolution from task automation to strategic augmentation, showcasing the transformative potential of AI in the modern workplace. These co-pilots operate in parallel with human workers to administer more precise work seamlessly.”

Kroupenev noted that the technology is “already showing impressive results for highly specific use cases, like AI-driven solutions that allow factory engineers to ensure production uptime, safety and sustainability through predictive maintenance. The evolution of AI co-pilots is set to make them ubiquitous across all sectors, transforming the workplace by empowering workers with data-driven insights into automation.”

As with digital twins, there are differing definitions of co-pilot tech that could well create some confusion. Kroupenev said that he is seeing a range of views on what constitutes a co-pilot and agrees that industry needs a tighter definition. He offered three components he believes to represent the core premise of co-pilots: 

  • Collaboration: AI co-pilots should be designed to augment and empower human work, making it more efficient and insightful through ongoing collaboration and interaction.  
  • Broad insight and applicability: AI co-pilots should be useful at multiple levels of work, including strategy, planning and execution, and they should be customizable for multiple applications. 
  • Agency: AI co-pilots should evolve to have agency and become more like a skillful team member than a tool to assist, collaborate and perform a wide range of tasks independently.   

“These concepts should remain generally constant even as the application of co-pilots can vary significantly across different fields,” he added.  

A tool for everyone?

Kroupenev believes that every worker will use an AI-copilot within the next five years due to AI’s increasing ability to aid the future workforce in conducting research, creating and executing action plans, summarizing and generating content, writing code, expediting planning processes and analyzing large amounts of data from multiple sources.

“On the factory floor, co-pilots built on a combination of multiple AI systems and generative AI (genAI) interfaces can leverage multiple real-time signals form the production process and text-based enterprise data, such as field notes and operating procedures,” he said. “This will empower managers to better resolve issues and allocate resources, engineers to plan production and production workers to speak to co-pilots in their natural language to learn when a machine needs repairs and receive step-by-step guidance on best practices for repairing the equipment. This will drive efficiency, innovation and sustainability in manufacturing and create opportunities for those with more diverse skill sets to start their careers in industry.”

The ongoing shortage of skilled labor for the manufacturing industries means that AI technologies will not just be useful to industry, but “crucial for enabling U.S. and European manufacturing to continue to grow,” Kroupenev said. “The addition of AI co-pilots will help reduce production downtime, meet sustainability goals, upskill workers, enable decision-making and more. The co-pilots will deeply understand topics across the industry, providing workers with the information they need to make informed decisions more autonomously.”

Evaluating the co-pilots

There’s been a lot of news about forthcoming AI co-pilot technology, for example from Siemens and Microsoft. But that’s just the tip of the iceberg. Kroupenev noted that there’s already several marketing-level showcases of early stage genAI-based applications in the industrial space. 

“Companies like Cognite, Siemens and C3 all have come out with co-pilot concepts early in the hype cycle,” he said, “but there is still some way to go to make these truly useful and reliable. The industrial world is heavily biased towards trust and reliability for a good reason, and making a co-pilot that can be fully trusted to provide accurate information and perform well in a manufacturing setting is currently one of the biggest obstacles to wider development and adoption. Reliable and unbiased data is what determines the accuracy of models, and manufacturing data is often neither reliable nor standardized.”

Kroupenev said Augury’s Machine Health technology is well positioned as a data source for co-pilots due to its focus on having a highly standardized data set and guaranteed accuracy of more than 99.9% for Augury’s AI-driven diagnostics. 

For manufacturers looking to further investigate AI-copilot technologies, Kroupenev said it’s crucial for manufacturers looking into this technology to first understand the specific problems that exist in their operations. 

This is a key lesson to apply to any technology—even tech that is more straightforward in its application, such as robots. As Ira Moskowitz, CEO of the ARM Institute noted in a recent Automation World podcast discussion: “One of the most common things you'll see is a robot arm sitting in the corner somewhere because the manufacturer thought they needed it and they bought it. And now it's sitting in the corner. Because of this, one of the first things manufacturers need to do is to get help triaging what their problem is. They may think it's an automation problem or a robotics problem, but it may not be. They may have other inefficiencies that need to be dealt with that sound like it requires a robot, but the root cause could be something else. So, the very first thing to do is an engineering analysis of the line and verify if you'll need a robot.”

Kroupenev sees a similar line of thinking being applied to AI. “A common misconception is that AI will solve all problems, but that’s not completely true,” he said. “AI has the potential to solve a lot of organizational obstacles, but each solution is unique in the way it works to solve the biggest challenges. When shopping around for AI solutions, asking questions is key. Understanding the quality of data, the types of insights generated and how the value is measured are critical before signing an agreement with an AI vendor. For the best outcomes, start with proven AI technologies that provide near-immediate value to gain buy-in for larger initiatives.”

Sponsored Recommendations

Wireless Data Acquisition System Case Studies

Wireless data acquisition systems are vital elements of connected factories, collecting data that allows operators to remotely access and visualize equipment and process information...

Strategizing for sustainable success in material handling and packaging

Download our visual factory brochure to explore how, together, we can fully optimize your industrial operations for ongoing success in material handling and packaging. As your...

A closer look at modern design considerations for food and beverage

With new and changing safety and hygiene regulations at top of mind, its easy to understand how other crucial aspects of machine design can get pushed aside. Our whitepaper explores...

Fueling the Future of Commercial EV Charging Infrastructure

Miguel Gudino, an Associate Application Engineer at RS, addresses various EV charging challenges and opportunities, ranging from charging station design strategies to the advanced...