3 fundamental steps to AI in manufacturing


A tradition of innovation


The manufacturing industry has traditionally been a driver for innovation, bringing new concepts to life in new products that drive our nation forward.


Today, many innovations begin in the virtual world — and that includes the evolving capabilities of artificial intelligence (AI). However, manufacturers can and should play a critical role in adopting and advancing AI capabilities. AI capabilities can enhance manufacturing from the shop floor to the top floor, and many leaders have recognized that now is the time to look at what AI can do.


The power of AI can be either good or bad — a poorly executed AI solution can quickly lead to misguided decisions or unexpected outcomes that put your business at risk. To successfully integrate AI, you need a foundation of understanding and a strategic approach.


Futurist Jim Carroll on AI in manufacturing



4:33 | Transcript




A foundation of intelligence in AI


AI is only as good as its digital composition — its algorithms, data and surrounding systems — along with the context where all of those elements are applied. So, you need to start by building an understanding. Large language models like ChatGPT are one type of AI technology, but there is much more to consider.

Tony Dinola

“You can't start considering an AI solution without understanding your business requirements and the technologies on the market.”

Tony Dinola

Grant Thornton Technology Modernization Partner


As a baseline, manufacturers need a clear understanding of the types of AI, the wide range of capabilities, the packaged solutions that are available and, most importantly, how AI can solve the problems and needs in their businesses. “You can't start considering an AI solution without understanding your business requirements and the technologies on the market,” said Grant Thornton Technology Modernization Partner Tony Dinola.


“There are a lot of opportunities that spin through my head when we talk about AI in manufacturing, because you can apply it to the shop floor, you can apply it to supply chain, you can apply it to quality,” said Grant Thornton Manufacturing National Managing Principal Robert Hersh. “The capabilities that we would consider in the finance organization, for instance, are different from the supply chain organization — they’re different from plant management, or even the sales organization.” The wide range of opportunities can seem intimidating or overwhelming if you don’t understand how, when and why your organization should focus on using AI technology.


To start building your understanding, and a strategy for success, think in terms of three fundamental steps:


Pick your problem

Robert Hersh

“It’s all about data. How do you ensure the data quality and data governance? How do you ensure that you are collecting the right data and using it in the right way?”

Robert Hersh

Grant Thornton Manufacturing Industry National Managing Principal

There's a lot of pressure to start exploring AI. But make sure you match the tool to the problem. How, where and why do you need AI?


To answer this question, you need a fundamental understanding of both AI capabilities and your business problems. Then, you need to identify the right problem and the best solution. “If you're going to apply this technology, how are you solving that problem? What does success look like?" asked Hersh. As you apply AI, you need to maintain a grasp of both the technology and the target — the process, program, problem or need that the technology is trying to address. But, beneath the surface, there’s a factor that must feed your AI solution: Data.


“It’s all about data,” Hersh said. “How do you ensure the data quality and data governance? How do you ensure that you are collecting the right data and using it in the right way?”


Whether you are going to solve your problem with a third-party product or a program that you developed internally, your AI solution will need to feed on data and its results will reflect the quality of that data.


Insight on this issue from other industries












Establish your governance


Any AI solution has the potential to influence a company’s processes, people, policies and decisions. So, you will need high-level governance even as you select the problem that you want to address. Once you identify your problem, and lay out a path paved with quality data, it’s time to fully establish the governance that will guide a successful outcome.


Good governance must continue to evolve as your AI implementations expand. “You will need general governance within the organization, in regards to how you deploy AI tools across the various use cases,” Dinola said. “You need to ensure deployment is being done in a controlled and intentional fashion. You also need to ensure a holistic approach to data management, master data management and master data governance, coupled with ethical considerations and risk mitigation.”


If an AI solution will have access to employee or customer data, for instance, you need to understand and govern the possible implications. You need to notify people about how their data is being used, as is sometimes required, and you need to examine various possible outcomes so that you can define applicable guidance and policies needed.


Your governance will rely heavily on your understanding of your existing policies and procedures, and the accuracy of your performance metrics. You need to carefully evaluate your current state, risks and opportunities as you define your governance and your strategy for success. You also need to review, align with, and possibly revise your related compliance programs to address the new risks posed by a new AI solution.


Define your success


A manufacturer’s AI initiative can come in many shapes and sizes. If you need better forecasting, you might seek an AI solution to help predict equipment failures and plan your maintenance, parts and procurement. Or, you might need forecasts of customer buying, supply chain availability, warehousing needs, new product demand and other planning factors. AI solutions in your line can help improve productivity and quality, while AI data analysis can process masses of data from remote sensors and unstructured documents to reach conclusions that were hidden before. Whatever the solution, you need a shared vision of the desired outcome. 

Tony Dinola

“You need to ensure that you're building a scalable solution which can change with the market and the ever-evolving technology.”

Tony Dinola

Grant Thornton Technology Modernization Partner


You need to clearly define success in a way that not only captures the expected goal, but also indicates when the effort is off target or needs modification. Ultimately, you will also need to understand how your solution can evolve over time — because your solution will need to change as your needs change and technologies evolve.


“You need to ensure that you're building a scalable solution which can change with the market and the ever-evolving technology. You need effective testing protocols, and then ultimately you need to evaluate the return on the output of the product so that you can continue to enhance it,” explained Dinola. 




Promising potential


In manufacturing, AI technology is already helping to monitor, measure and enhance the performance of machines, analyzing supply chain data to help optimize inventory levels and much more. Yet, it is still in its early stages, when you consider the host of opportunities and unknowns. Now might be the time for you to consider AI capabilities, but it’s important to be realistic and strategic about how you apply them. 


As manufacturers continue to move toward integrating AI into machinery, devices, processes, programs and more, they will need AI strategies with sound governance to harness all of the promising potential.




Get the outlook for other key industries




Our manufacturing featured industry insights