AI Can Revolutionise Manufacturing Efficiency

Adapting new technologies to manufacturing applications and processes is often a challenge – and, particularly with artificial intelligence (AI), as many manufacturers are beginning to discover, there is no easy button to push. The complexities associated with AI and the scarcity of AI experts have kept its potential benefits out of reach for many manufacturers. Those that have successfully blazed the AI path have had to leverage highly individualised tools, which are difficult to fund and execute, and may ultimately offer lower returns compared to other industries.

Still, the benefits of AI are undeniable, and manufacturers continue to find new AI use cases to improve efficiency and customer service. In fact, according to Protiviti’s latest global survey of CFOs and finance leaders, among finance organisations in manufacturing and distribution that are employing generative AI, 71% are using it for process automation and 53% are doing so for financial forecasting.

For manufacturers, making this bold leap requires understanding the benefits of AI applications and mastering a sound approach to adoption. These organisations also need to consider how AI tools should be governed to ensure that they are designed, developed, deployed and used responsibly.

The state of AI play

A number of real-life use cases in areas such as decision support, customer experience and knowledge management support the value AI can deliver today.

Decision support tools – In manufacturing and distribution organisations, AI decision-support applications are being deployed to curate and synthesise large amounts of data from the enterprise’s own information and operations technology assets. By presenting that data in meaningful contexts, AI decision support applications create the opportunity to make better-informed decisions.

For example, when COVID-19 vaccines became available a few years ago, one pharmacy chain sought to understand where and when the vaccines would be needed the most. The organisation (with Protiviti’s assistance) was able to deliver an AI-driven model to project demand for — and optimise distribution to — long-term care facilities. Using machine learning, the application predicted demand with over 90% accuracy.

Customer experience applications – AI customer experience applications increase response times, address customer service needs, generate personalised experiences and make customer interactions more convenient.

Consider this example: A car manufacturer deployed generative AI to make voice control in its vehicles more intuitive. While conventional voice control is constrained to specific, pre-coded tasks and responses, large language models expand the topics and tasks that voice control systems can handle.

Knowledge management applications – With AI knowledge management applications, manufacturers can make expertise already present easily accessible and interpretable across the enterprise. Massive amounts of data can be reorganised into a meaningful synopsis.

To illustrate, a global manufacturing company was struggling with transoceanic freight costs. The company, which had trouble anticipating cost changes and was experiencing significant volatility as a result, deployed an AI application to predict port-to-port journey costs by month. The approach accounted for previously unforeseen trends to contain the cost of freight.

Getting off the ground with a pilot

Knowing when and how to start the AI journey is critical. Manufacturers can start this process by first formulating their AI vision and recruiting a cross-functional solution team to look at all aspects of the AI proposal and design a governance framework accordingly.

A few important points to consider: First, the cross-functional team should look at the best use cases to maximise the pilot’s return on investment (ROI). Second, while AI governance is an important consideration, the good news is that in a manufacturing environment, there are many potential use cases that are lower risk in regard to areas such as compliance, privacy, transparency and bias, compared with use cases in other industries (e.g., financial services) that must address higher levels of risk in these areas. This means that the level of AI governance risk in manufacturing is lower while the potential ROI from successful use cases, as noted earlier, could be higher compared with other industries.

Next, the team should conduct a pilot of the AI model. When it comes to pilots, here are a few important things to note:

• Many pilots result in production-ready applications.
• All pilots teach invaluable AI implementation skills.
• Pilots reveal where teams need to learn more or how to augment their skills accordingly.

Two other important points:

• The AI vision – As noted above, it’s important to first articulate a purpose for the AI effort. The purpose could be decision support, customer experience, knowledge management, or automation and process efficiencies. After articulating a purpose and potential ROI, the team can then identify a suitable sponsor for the effort and tap resources to assemble a cross-functional team.
• The AI team – Leaders should also consider not only technology and operations resources but also risk management, change management, and privacy and security skills. Ideally, the team will have free rein to evaluate and experiment with the technology. As the pilot evolves toward a fully fledged production application, leaders can recalibrate its oversight.

May the best AI idea win

In terms of the technology itself, the newly formed team should explore which AI applications other manufacturers or competitors are using. The goal here is not to copy what others are doing but rather to identify where AI can make dramatic gains or even disrupt current models. Teams should generate opportunities or ideas that align to the vision and have the biggest potential to deliver significant value.

Prioritising ideas – Teams can assess the complexity and feasibility of each idea they’ve generated while also considering the technologies available. Members should analyse requirements, assess data integrations and consider operational readiness for each opportunity. They should also weigh the financial benefits against complexity to deliver a single, well-qualified pilot recommendation.

After delivering the pilot, it’s important for the team to evaluate both the process and the product. If necessary, the team should refine its methods and consider how to improve the application. Consider, for example, the sustainability of the solution. Is the process or technology going to change substantially in the near future, thus potentially requiring changes to the AI solution? The team also may revisit the AI ideas they have already developed as the basis for further AI projects.

Key takeaways

While adopting innovative technologies can be especially challenging for manufacturing enterprises, they can reduce implementation pain points by following the recommendations outlined here. For manufacturers, there is probably more upside than risk because there are many potentially high-return use cases where they don’t face the same compliance or privacy risks as do organisations in other industries.

Lastly, remember that with AI, getting it right is better than moving quickly. Do not forget these key takeaways:

• Invest time and resources to understand the organisation’s opportunities.
• Put together a cross-functional team to create and master a sound AI development approach.
• Keep in mind that in addition to direct benefits and ROI, pilots are particularly helpful for increasing team confidence and expertise while highlighting opportunities for further skills development.