Accuracy concerns slow manufacturers’ Gen AI adoption

Manufacturers are implementing generative AI initiatives slower than anticipated due to accuracy concerns, according to a report from Lucidworks.

The study surveyed over 2,500 global AI decision-makers and found that 58% of manufacturing leaders plan to increase AI spending in 2024, down from 93% in 2023.

Additionally, 44% of manufacturers have concerns about response accuracy, while only 3% are worried about job displacement.

“Across the board, concerns around security, response accuracy, and costs have forced most businesses to slow down their planned initiatives and be more strategic about the balance between cost and benefit,” Lucidworks said in a statement. “Security worries have tripled, accuracy concerns have grown fivefold, and transparency issues have quadrupled since 2023.”

Challenges in the manufacturing sector

The reluctance to embrace Gen AI in manufacturing isn’t quite surprising to analysts. “Data security and response accuracy concerns are particularly acute in the manufacturing sector due to the high stakes in production processes and the proprietary nature of manufacturing data,” said Thomas George, president of Cybermedia Research. “Compared to other industries, manufacturers must be more cautious in protecting their information and ensuring correct results from AI automation outputs that govern process integrity and product quality.”

Manufacturers also face technical and operational challenges, such as the need for retrofitting existing systems, that contribute to their hesitation in adopting Gen AI.

“The most prominent challenge in integrating generative AI into existing manufacturing systems is the complexity of legacy infrastructures, which necessitates considerable customization,” George added. “Manufacturers also must grapple with data quality concerns and whether existing data resources are sufficient for training these AI models well enough.” 

Commercial vs opensource

Lucidworks also found that 47% of companies use commercial LLMs like Gemini and ChatGPT alone, while 30% have opted for open source exclusively. The rest use a mix of both.

George pointed out that while commercial LLMs offer better support, the costs can be prohibitive for some organizations.

“Commercial LLMs offer more straightforward integration mechanisms, ongoing support, and robust performance guarantees in industrial cases where consequences are severe,” George said. “However, such technologies may come at higher costs and have fewer options than open-source ones. Open-source models give flexibility options, including cost benefits – but they require a firm knowledge of how they should be managed internally.”

This choice deeply affects a manufacturer’s competitiveness, George added. Commercial solutions tend to provide quicker implementation, leading to potentially faster time-to-value. On the other hand, open-source models may allow more customization for specific needs.