The AI questions manufacturers should be asking

Executives and boards at manufacturers are asking questions about AI initiatives. Often, they ask about capabilities such as computer vision, natural language processing and digital twins.

The better place to start, however, is with business needs. “If there is one question leaders should be asking about AI, then it is definitely, ‘How will this make us more resilient?’” said Grant Thornton AI & Data Partner Sumeet Mahajan. Manufacturing resilience sits on a triad:

• Throughput
• Quality
• Cost control

Some companies also depend upon product development and other factors, but every manufacturer’s resilience requires a balance of these three elements.

How can AI help you improve throughput and elevate quality while controlling costs? That leads to some next-level questions about AI.

Next-level questions

Think about AI in terms of: Your business needs + How does AI offer a solution we didn’t have before? You can examine this through a lens of throughput, quality and cost control.

Throughput questions

Quality questions

Cost control questions

At this level, many of the questions are familiar. These are questions that many manufacturers have asked, but AI is changing the answers with new capabilities.

These new AI capabilities are already helping companies manage the growing volatility of unexpected events. As volatility continues, resilience will depend on agility. Manufacturers need AI that is agile enough to answer changing questions that become increasingly granular, reaching every level of the business.

Long-term business questions might remain familiar, but short-term questions evolve quickly.

When manufacturers had more stability in their supply chains, distribution and markets, they could use strategies built upon just-in-time manufacturing and distribution. Ever-increasing disruptions and uncertainties have made production cycles less dependable, so now manufacturers need agility that can adapt at every level. 

Proactive adaptation

AI can help manufacturers adapt proactively, before external factors drive a change. “You need to have some intentional flexibility in your systems,” Mahajan said. “That could mean something as simple as maintaining safety stock, or it could mean that you don't only use the lowest-cost supplier for a part.” 

“Resiliency and flexibility don’t always relate to negative events; they can also relate to shifting customer demands and how you capture more market during that shift,” Mahajan said.

As these questions become more time-specific, they can also become more function-specific. A manufacturer’s agility and resilience are increasingly defined by its ability to find the right answers and make quick decisions at every level.

“If we look at function-by-function examples of an organization, we can find a view of resiliency and flexibility for every subfunction,” Mahajan said. “If you look at opportunities from supply chain operations, a manufacturing standpoint, a logistics standpoint or a commercial standpoint, you’ll come up ways to build resiliency and flexibility.”

The questions that manufacturers ask about AI are changing and becoming more pervasive throughout the organization. That’s why many companies need to think beyond getting answers once and toward an ongoing capability to find answers with AI-driven scenario planning.

How we can help you

INDUSTRY

Manufacturing -->

Ready to talk? We’re ready to listen.

Request a meeting -->

To capture the answers for increasingly complex questions that continue to evolve, some manufacturers are developing and using an AI scenario-planning capability. “A lot of planning and building for resiliency has to do with scenario planning,” Mahajan said. “The capability of scenario planning is not new, but it has matured with AI.”

With scenario planning and AI capabilities at multiple levels, manufacturers can shift from answering questions to automating the insights they will need.

Even an advanced analysis is not as powerful as a solution that pairs analysis with action. “In the world before AI agents, the best-in-class predictive maintenance solutions would use a machine learning model based on a machine’s data to predict when a failure was going to happen,” Mahajan said. “Then, someone would have to manually look at the report, schedule an engineer and then the engineer would place an order for replacement part.”

“Now, your prediction of when the machine is going down is a lot more accurate — because you’re using forward-played synthetic scenarios, and you're also prescribing what the most likely machine breakdown is going to be,” Mahajan said. “The agent automatically creates a work order for an engineer and a work order for a replacement part.”

“You can get to a level of accuracy that is much higher because it’s more comprehensive,” Mahajan said. “That puts you in a better position to say, ‘Even if a bigger order comes in, the system is resilient enough to deliver, but we need to do maintenance on this Saturday, during second shift.’”

Can you process the next big order that comes in? How can AI help you prepare for it? Those are the kinds of questions you should be asking. The answers will drive your resilience.