[Manufacturing x AI Utilization Barriers] Over 40% Cite 'Insufficient Training Data' as the Biggest Challenge in AI Adoption; Nearly Half Say Investment Should Prioritize 'Data Collection Infrastructure' Over 'AI Model Implementation'

Simtops, Inc., provider of the No. 1 domestic market share on-site reporting system 'i-Reporter,' conducted a survey on the importance of data preparation in AI utilization in the manufacturing industry, targeting 111 DX and AI promotion managers.
The survey revealed that while about 90% of companies have begun utilizing AI, 87.4% consider 'organizing and structuring primary on-site information' more important than 'selecting AI models or tools.' In terms of areas to invest in over the next three years, 'development of data collection infrastructure' (47.7%) surpassed 'AI model implementation' (38.7%) to take the top spot, suggesting that the success of manufacturing in the AI era depends on solid data preparation.

■About 90% of DX/AI promotion managers are utilizing AI in 'full implementation' or 'PoC stage'
When asked, 'Q1. Which of the following best describes the progress of AI utilization at your company?' (n=111), 45.0% answered 'Fully implemented and used in operations,' and 44.1% answered 'Working on it at the PoC (Proof of Concept) stage.'

■95.9% of those utilizing AI feel they are achieving 'sufficient' or 'some' expected results
To those who answered 'Fully implemented' or 'PoC stage' in Q1, we asked, 'Q2. Is your company's AI utilization achieving the initially expected results?' (n=99). 32.3% said 'Sufficiently achieved,' and 63.6% said 'Achieved to some extent.'

■'Insufficient training data' is the top challenge for AI utilization, cited by 44.1% of DX/AI managers
When asked, 'Q3. What challenges do you feel in promoting (or considering) AI utilization at your company? (multiple answers)' (n=111), 'Insufficient amount of data for AI training' was at 44.1%, 'Shortage of internal personnel to promote AI utilization' at 42.3%, and 'Data has many omissions or errors' at 38.7%.

■About half report that '50% to less than 80%' of primary on-site information is digitized and structured
When asked, 'Q4. What percentage of primary on-site information (work records, inspection records, etc.) at your company do you think is digitized and structured in a way that AI can use?' (n=111), 49.5% answered '50% to less than 80%,' and 23.4% answered '30% to less than 50%.'

■87.4% of DX/AI managers prioritize 'organizing and structuring primary on-site information' over 'selecting AI models/tools'
When asked, 'Q5. Do you think it is more important to organize and structure primary on-site information than to select AI models or tools in manufacturing AI utilization?' (n=111), 26.1% answered 'Strongly agree,' and 61.3% answered 'Somewhat agree.'

■About 90% of managers recognize that 'structuring primary on-site information is a prerequisite for AI utilization'
When asked, 'Q6. Do you think structuring primary on-site information is a prerequisite for AI utilization in manufacturing?' (n=111), 36.0% answered 'Strongly agree,' and 50.5% answered 'Somewhat agree.'

■Top reason for considering structuring primary information a prerequisite: 'On-site judgment know-how can only be extracted from primary information' at 65.6%
To those who answered 'Strongly agree' or 'Somewhat agree' in Q6, we asked, 'Q7. Why do you think structuring primary on-site information is a prerequisite for AI utilization in manufacturing? (multiple answers)' (n=96). 'Because on-site judgment know-how can only be extracted from primary information' was at 65.6%, 'Because analysis is difficult with unstructured data' at 49.0%, and 'Because AI cannot read handwritten or verbal information' at 42.7%.

■Top areas for investment in the next 3 years: 'Development of data collection infrastructure' (47.7%) and 'Improvement of data quality/cleansing' (41.4%)
When asked, 'Q8. What areas should your company particularly invest in over the next three years to succeed in AI utilization? (up to 3 answers)' (n=111), 'Development of data collection infrastructure' was at 47.7%, 'Improvement of data quality and cleansing' at 41.4%, and 'Introduction of AI models or tools' at 38.7%.

■Over 60% of DX/AI managers have introduced a 'dedicated data collection system' for structuring primary information
When asked, 'Q9. What initiatives is your company currently undertaking to structure primary on-site information? (multiple answers)' (n=111), 'Introduced a dedicated data collection system' was at 61.3%, 'Consolidating on-site data with Excel, etc.' at 42.3%, and 'Preparing data formats for AI' at 35.1%.

■Summary
This time, we conducted a survey on the importance of data preparation in AI utilization in manufacturing, targeting 111 DX and AI promotion managers. The results showed that while about 90% have started utilizing AI, 87.4% consider 'organizing and structuring primary on-site information' more important than selecting AI models or tools.
The survey revealed that while the adoption rate and perceived results of AI in manufacturing are high, the整備 of the data infrastructure that supports it remains a major bottleneck. Strategically investing in the digitization of on-site reports like work, inspection, and check records, and building a data collection infrastructure in parallel with advancing AI models, will be key to sustainably expanding the results of AI utilization.