The Primary AI Question For Manufacturers

The Pertinent Question​

I believe manufacturers should not begin their technology transformation journey by asking when or how to introduce AI. They should begin by asking a much more fundamental question: What problems are we trying to solve?

That distinction matters.

Technology transformation should not become a technology-led exercise simply because a technology is new, powerful or receiving enormous attention. AI is a means to an outcome—not the outcome itself. Sometimes AI may be the right answer. Sometimes analytics, automation, better data engineering, cloud modernization, process redesign or a combination of technologies may be the better answer.

The starting point, therefore, should be the business problem.​

Real Business Considerations​

Consider a manufacturer struggling with unplanned equipment downtime. The conversation should not begin with, “Where can we deploy AI?” It should begin with, “Why are we experiencing downtime, what data do we have, how quickly can we identify the causes and what would a meaningful reduction in downtime be worth?”

The eventual solution could involve predictive analytics, machine learning, IoT data, better data pipelines, a modern data platform, workflow automation or several of these together. AI may be an important component, but it is the problem statement that determines the technology, not the other way around.​

This way of thinking is increasingly reflected in industry research. Gartner’s research on AI use cases in manufacturing emphasizes the importance of selecting the right AI use cases based on business value and feasibility rather than treating AI adoption as an end in itself. That reinforces something I have come to believe: the question is not how much AI a manufacturer can introduce, but where technology can create measurable value.

That leads me to a simple framework for manufacturers.

The Manufacturing Framework​

First, identify the business problem. Is the organization trying to improve yield? Reduce scrap? Increase throughput? Lower energy consumption? Improve forecast accuracy? Reduce inventory? Shorten changeover time? Improve quality? Reduce unplanned downtime? Accelerate engineering? Improve customer service?

Second, quantify the problem. A problem becomes a transformation opportunity when we understand its business impact. If a production line loses four hours a week because of recurring failures, what does that mean in lost production, labor, energy and customer commitments? If inventory is consistently higher than necessary, what is the working-capital impact? If engineers spend hours searching for previous designs, what is the opportunity cost?

Third, understand the data and technology landscape. McKinsey’s research on data-quality roadblocks in manufacturing highlights how fragmented systems, data-quality challenges and legacy environments can prevent manufacturers from realizing the value of AI. This is why I believe data readiness must be treated as part of the transformation journey—not as an afterthought.

Only then should we ask: Which technology can solve this problem most effectively?

The Answers Will Vary​

Sometimes the answer will be AI. Sometimes it will be a modern data platform. Sometimes it will be automation. Sometimes it will be cloud modernization. And increasingly, it will be a combination of technologies working together.

I also believe this approach changes the conversation between technology leaders and business leaders. Instead of presenting a catalogue of AI capabilities—computer vision, copilots, predictive models, digital twins or agents—we can start with business outcomes and work backward to the architecture and technology required to achieve them.

This is especially important in manufacturing because the technology landscape is rarely greenfield. Plants have operational technology, legacy systems, machines of different generations, disconnected data sources and processes that have evolved over decades. A successful transformation must respect that reality.

McKinsey’s work on connecting use cases in Industrial IoT transformation makes a related point: manufacturers can create more value by identifying meaningful use cases and connecting them through a scalable technology architecture, rather than allowing transformation to become a collection of isolated experiments. For me, this is a critical distinction between adopting technology and transforming a business.

There is another reason I would resist starting with “When should we start with AI?” The question can create a false sense of urgency. It can make organizations feel that they are behind simply because another company has announced an AI initiative.

The Final Question​

The better question is: Where is the business value today, and what is preventing us from capturing it?

If AI can unlock that value, start there. If the organization first needs to modernize its data foundation, start there. If the process itself is broken, fix the process. If people need better decision support, build it. If the opportunity is too small to justify the investment, do not force AI into it.

For me, this is the mindset shift manufacturing needs: from technology-first transformation to problem-first transformation.

AI will undoubtedly play a significant role in the future of manufacturing. I am not arguing against AI adoption. I am arguing for better questions before adoption.

The manufacturers that create lasting value will not necessarily be the ones that introduce the most AI. They will be the ones that become very good at identifying important problems, understanding their economics, building the right data and technology foundations and applying the right tools to solve them.

So, when someone asks me, “When should we start with AI?” my answer is becoming simpler: Start when you have a problem worth solving—and let the problem tell you whether AI is the answer.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?