A recent survey by Edge Impulse and Manufacturing Dive’s studioID found that the majority of manufacturers are well on their way to adopting AI on embedded devices. Known as edge AI, this type of machine learning (ML) occurs on network endpoints rather than being uploaded to a cloud or server. The ability to quickly and reliably generate insights is ripe for innovation, enabling manufacturers to improve workflows, streamline processes and control costs.

Overall, the survey uncovered three notable findings: 

1. Production quality is a top concern. 

Nine in 10 manufacturers in the general market sample reported that they monitor production output, product quality and consistency and equipment. This aligns closely with where manufacturers are already investing in monitoring sensors and technologies that can detect anomalies and implement predictive maintenance.

Manufacturers expect that AI and ML will yield a number of operational or production efficiencies. From the general market sample, companies are looking for AI to improve accuracy (64%), automation (63%), safety (58%), production downtime (53%) and maintenance (52%).

“The goal is to reduce the need for human supervision,” says Jim Bruges, staff solutions engineer for Edge Impulse. “Many of our customers are working to train models that combine human intelligence with the repeatability of a machine. This reliability brings increased safety, reduced machinery downtime and reduced production costs.”

2. Real-time insights from AI are key to strategic improvements.

Considering the vast number of internet-connected sensors used throughout a manufacturing plant, edge AI is well suited to analyzing this stream of granular data. Manufacturers can leverage this deeper understanding of their production environment to resolve minor issues before they become larger problems. 

To do so, manufacturers are focusing on monitoring the factors most associated with equipment performance and manufacturing processes. For example, over 50% of respondents from the general market sample are measuring specific readings like temperature and air quality. Humidity and noise are also tracked by around 45% of respondents.

3. Internal expertise for AI is a competitive advantage.

As manufacturers move forward with AI implementation, they must consider which role or department should lead this initiative. This is crucial as one of the barriers to adoption was a lack of staff expertise, which was cited by 36% of the general market group and 48% of the Edge Impulse customer sample. 

This is further reinforced by the survey findings that show while many companies have a member of the C-suite team responsible for AI adoption, less than 10% of either survey groups have dedicated AI teams. This indicates that internal domain expertise will be a critical factor.

“For successful edge AI integrations, it’s crucial for manufacturers to recognize that AI goes beyond just the technology,” Daniel Situnayake, Director of AL/ML for Edge Impulse says. “Understanding the full ML Ops process, from data collection and model training to deployment and continuous monitoring, is key, and can vary widely depending on specific use cases.” 

Survey Methodology 

Edge Impulse and Manufacturing Dive’s studioID surveyed 150 manufacturing leaders that are already implementing AI and ML into their production environments. This general market sample represents various market segments in the U.S., U.K., and Canada, including heavy industrial (33%), technology/electronics (25%) and customer packaged goods (19%). The majority are contract manufacturers (60%). 

Additionally, Edge Impulse surveyed more than 130 of its global customers who also bring expertise in the technical implementation of AI/ML. This customer sample primarily employs less than 1,000 people (93%), and the majority are not contract manufacturers (78%).