The Future of AI in Manufacturing
Thejeswi Ramesh
NOV 21, 2024
From 1979 to 2020, U.S. manufacturing jobs dropped by 36%, while all other non-farm jobs in the U.S. grew by 65%. A significant portion of critical manufacturing moved to China, leaving only simple assembly of finished goods in the U.S. Since skilled labor in China was inexpensive, there was little pressure to adopt advanced technologies. However, during this period, we saw substantial growth in key manufacturing technologies. For instance, PLC and SCADA systems grew into a $22 billion market. The robotic arms industry also experienced massive growth, now a $36 billion market, while legacy machine vision systems expanded into an $18 billion market.
However, things are changing rapidly as factories return to the U.S. Since 2020, manufacturing construction spending in the U.S. has surged by 106%. But who will run these factories? The U.S. faces a massive shortage of skilled labor: 82% of manufacturing companies are experiencing labor shortages, and this problem is only set to worsen. This shift makes digital technology a necessity rather than a luxury. Instead of paying $10,000 for labor in China, companies now have to shell out $60,000–$70,000 for U.S. labor. In this transition, investing $15,000–$20,000 in a digital equivalent is a prudent solution.
Manufacturers today allocate 39% of their factory budgets to digital technologies. While less than 12% have implemented comprehensive smart manufacturing initiatives, 80% are transitioning toward full adoption. According to Accenture and UBS, AI will unlock 45% of gross value added in manufacturing—higher than any other industry—potentially growing into a $700 billion market by 2030.
Legacy brands like Cognex, Keyence, and Omron dominate the $18 billion machine vision market. Cognex, listed in the U.S., is a pioneer in machine vision technology, while Keyence and Omron, Japanese-listed companies, have a significant presence. Their rule-based systems achieve 80–85% accuracy in sectors such as automotive and 90–95% in more controlled environments like electronics and semiconductors. For context, human inspection accuracy averages around 90%, making these systems both less accurate and more costly than manual inspection in many regions. These systems also fail to cover all inspection parameters, meaning manufacturers still need human resources to handle the remaining inspections and verify the systems’ results.
Several companies worldwide have emerged, developing advanced AI technologies for visual inspection in manufacturing. However, we faced three main challenges in finding a winner to back in this space:
- Service-Centric Models: Most companies in this space are service-oriented, custom-developing solutions on a case-by-case basis. Feedback from the former head of manufacturing at Nvidia, who also led manufacturing for Amazon Alexa, highlighted this issue. The service model is not scalable and has been a significant barrier to growth. Custom development for every client limits scalability and hinders adoption for end customers.
- Accuracy Issues: Implementing AI in manufacturing is not straightforward. Many nuances must be studied carefully, and if a system doesn’t achieve 99%+ accuracy, it becomes redundant. Manufacturing environments demand extremely high precision, which many AI solutions fail to deliver.
- Feasibility Challenges: Solutions, even from marquee players like Landing AI, often struggle with feasibility due to the significant data and compute resources required for widespread adoption in manufacturing.
In our search for a company that addresses these challenges, we found Aditya, Dharmagya, Subhra, and the team at Frinks.ai. They offer a scalable, highly accurate, and feasible solution designed for real-world manufacturing environments.
Frinks enables manufacturers to seamlessly integrate advanced AI for quality control and visual inspection, ensuring unparalleled quality and consistency in manufacturing. Their vision AI framework is designed to enhance visual cognition in manufacturing and supports a wide range of use cases, including quality control, compliance monitoring, safety monitoring, object identification, inventory counting, measurements, positioning and guidance, predictive maintenance, and more.
Frinks is building foundational vision models with a unique training framework and proprietary vision algorithms to address the challenges of implementing AI in manufacturing. With this innovation, the company provides a written money-back guarantee of 99.99%+ accuracy to all its customers. Alongside the foundational AI framework, Frinks is also developing a no-code platform. The manufacturing industry’s diversity and high stakes make managing and scaling custom-developed software nearly impossible.
Recognizing these challenges, our no-code platform allows manufacturers to build visual inspection projects, train and deploy highly accurate, domain-specific custom vision AI models derived from our foundational AI framework, and customize workflows to meet unique project requirements. This ensures manufacturers can address their specific products and use cases without coding expertise. The platform’s sector-agnostic design enables horizontal scalability across industry verticals. Watching the team scale this company from a mere paper plan in 2023 to over 12 clients today in a profitable and accretive manner has been a joy. We look forward to their journey ahead and are more excited about innovation in the manufacturing space than ever!
Chiratae Sonic, our flagship seed Initiative is accepting applications till 30th Nov, 2024 at https://www.chiratae.com/chiratae-sonic/. We invite founders to reach out with details of their venture and receive initial funding reviews within just 48 hours for requests up to USD 2 million.