Driving Innovation in Manufacturing: How AI and the GenAI Center of Excellence are Driving Modernization

introduction

Technologies such as machine learning (ML), artificial intelligence (AI), and generative AI (GenAI) are ushering in a new era of efficient and sustainable manufacturing and empowering the workforce. Areas where AI can be applied in manufacturing include predictive maintenance, defect detection, supply chain visibility, demand forecasting, and product design. Benefits include increased uptime and safety, reduced waste and costs, improved operational efficiency, enhanced product and customer experience, and faster time to market. Many manufacturers are starting to adopt AI: Georgia-Pacific uses computer vision to reduce paper tears, improve quality, and add millions of dollars to profits; Baxter was able to prevent 500 hours of downtime in one facility alone with AI-powered predictive maintenance.

However, many enterprises struggle to fully leverage AI due to weak organizational and technology foundations (according to a recent World Economic Forum survey). Reasons include skills shortage, resistance to change, lack of quality data, and technology integration challenges. AI projects often get stuck in the pilot stage and never scale for production. Successfully leveraging AI and Gen AI technologies requires a holistic approach across cultural and organizational dimensions in addition to technical expertise. In this blog, we discuss how an AI Center of Excellence (AI CoE) offers a comprehensive approach to accelerate modernization through AI and Gen AI adoption.

Challenges of introducing AI into the manufacturing industry

Manufacturing faces unique challenges when it comes to adopting AI because it requires blending the traditional physical world (operational technology, OT) with the digital world (information technology, IT). Challenges include cultural norms, organizational structures, and technological constraints.

Factory workers deal with mission-critical OT systems. They prioritize uptime and safety and perceive change as risky. Cybersecurity was not a high priority because systems were isolated from the open internet. Traditional factory operators relied on experience gained through years of operational decision-making. Understanding how AI systems arrive at their decisions is essential to earning their trust and overcoming barriers to adoption. Factory teams are siloed, autonomous, and operate under local leadership, making AI adoption difficult. Initial investments in AI systems and infrastructure can be costly depending on the approach, and many manufacturers may struggle to justify the expense.

AI relies on large volumes of high-quality data, but in many manufacturing environments, this data can be fragmented, outdated, or inaccessible. Manufacturing legacy systems often run on vendor-dependent, proprietary software and use non-standard protocols and data formats, making integration with AI a challenge. Remote locations have limited internet connectivity, so manufacturing systems rely on accurate, reliable real-time responses and must overcome latency challenges. For example, AI systems must process sensor data and camera images in real time to identify defects as products move through the production line. A slight delay in detection could prevent a defective product from passing through quality control. In addition, manufacturing AI systems must meet strict regulatory requirements and industry standards, complicating the AI development and deployment process. The field of AI is still evolving and there is a lack of standardization of tools, frameworks, and methodologies.

Leadership Roles

Transformative AI adoption requires commitment and alignment from senior management in both OT and IT. OT leaders benefit from recognizing that connected, smart industrial operations simplify their work without compromising uptime, safety, security, or reliability. Similarly, IT leaders can unlock business value through AI technologies if they understand the uniqueness of manufacturing floor requirements. In fact, OT can be considered a business function enabled by IT. Integrating OT and IT perspectives is essential to realizing the business value of AI, including increased revenue, new products, and improved productivity. Leadership must develop a clear vision that connects AI to strategic objectives and foster a collaborative culture to drive functional and cultural change.

A vision tells the “why” of AI adoption, but successful AI adoption requires translating vision into action. An AI CoE bridges the gap between vision and action.

Accelerating AI Adoption and Business Outcomes with the AI CoE

Overview: An AI CoE is a multidisciplinary team of passionate AI and manufacturing subject matter experts (SMEs) driving responsible AI adoption. It promotes human-centric AI, standardizes best practices, provides expertise, upskills employees, and ensures governance. It develops a modernization roadmap focused on edge computing and modern data platforms. An AI CoE can start small with 2-4 members and scale as needed. To be successful, an AI COE requires executive sponsorship and the ability to act autonomously. Figure 1 outlines the core capabilities of an AI CoE.

Figure 1 Functions of the AI CoE

Explainable AI

The AI CoE must drive explainable AI in manufacturing, where safety and uptime are crucial. For example, if an AI model predicts equipment failure, a binary AI output such as “high probability of failure” or “low probability of failure” will not inspire the trust of factory personnel. Instead, an output such as “There is a 15% increase in vibration detected by the bearing sensor, which indicates a high probability of failure, which resembles historical bearing failure patterns” will make people more likely to trust the AI's advice. AWS offers multiple ways to increase the explainability of your AI models.

Enabling skills, building trust, and transparency

AI CoEs should work with HR and executives to upskill staff for an AI-enabled workplace by developing career paths and training programs that leverage existing skills. GenAI solutions can help close the skills gap by showing how AI complements employee expertise. Leaders should highlight how AI-enabled capabilities can save time spent solving complex problems and interpreting AI insights. For example, Hitachi, Ericsson, and AWS demonstrated computer vision using private 5G wireless networks that can simultaneously inspect 24 times more components than manual inspection to detect defects.

Work backwards from business outcomes, collaboration and breaking down silos

The AI CoE ensures collaboration and co-decision-making between AI solution builders and factory domain experts. They work together to work backwards from business objectives and break down silos to converge on AI solutions to achieve desired outcomes. Additionally, the CoE acts as a hub to identify impactful AI use cases by evaluating factors such as data availability, likelihood of rapid wins, and business value. For example, in a textile factory, the AI CoE can leverage data analytics to optimize energy-intensive processes to realize cost savings and sustainability benefits. Explore additional use cases in the AWS AI Use Case Explorer.

Governance and Data Platform

Governance and a data platform are essential to scaling manufacturing AI. The CoE will establish policies, standards, and processes for responsible, safe, and ethical use of AI, including data governance and model lifecycle management. AWS provides several tools to build and deploy AI solutions responsibly. The CoE will develop a secure data platform to connect disparate sources, enable real-time analytics, scalable AI, and achieve regulatory compliance. This data foundation will be the foundation for broader AI adoption. This is demonstrated by Merck's manufacturing data and analytics platform on AWS, which has tripled performance and reduced costs by 50%.

AI Technology, Tools and Automation

The AI CoE will evaluate and standardize AI and GenAI technologies, tools, and vendors based on manufacturing needs, requirements, and best practices. AWS offers a comprehensive set of AI and Gen AI services to build, deploy, and manage solutions that reinvent customer experiences. Scaling AI requires automation. The AI CoE will design an automated data and deployment pipeline that reduces manual work and errors and accelerates time to market. Toyota sets an example of large-scale AI deployment by using AWS services to process data from millions of vehicles and enable real-time response in emergency situations.

Measuring the effectiveness of a CoE

The value of an AI CoE needs to be measured from a business perspective. This requires a holistic approach that combines both hard and soft metrics. Metrics should include business outcomes such as ROI, improved customer experience, efficiency, and increased productivity of manufacturing operations. Internal surveys can measure employee and stakeholder sentiment toward AI. These metrics can help stakeholders understand the value of the AI CoE and their investment.

Start your AI CoE

Figure 2 Steps for building an AI CoE

Establishing an AI CoE requires a phased approach, as shown in Figure 2. The first step is to secure executive support from both OT and IT leadership. The next step is to assemble a diverse team of experts, consisting of frontline employees and AI IT experts. The team is AI trained and defines the CoE's objectives. They identify and deliver pilot use cases to demonstrate value. At the same time, they develop and strengthen the governance framework, provide training, foster collaboration, gather feedback, and iterate for continuous improvement. Integrating Gen AI further strengthens the content creation and problem-solving capabilities of the CoE, accelerating AI adoption across the enterprise. The AI CoE evolves over time. It starts out in a hands-on role, building expertise, setting standards, and initiating pilot projects. Over time, it transitions to an advisory role, providing training, fostering collaboration, and tracking success metrics. This empowers employees and ensures long-term AI adoption.

Conclusion

AI and GenAI technologies have the potential to create revolutionary new product designs, drive unprecedented levels of manufacturing productivity, and optimize supply chain applications. Adopting these technologies requires a holistic approach that addresses technical, organizational, and cultural challenges. The AI CoE serves as a catalyst to bridge the gap between business needs and responsible AI solutions. It fosters collaboration, training, and data solutions to optimize efficiency, reduce costs, and drive innovation on the factory floor.