AI in Manufacturing: Over Half Remain at the Pilot Stage. Why Connected AI Changes Productivity
The adoption of AI in the manufacturing industry is steadily expanding.
Visual inspection, predictive maintenance, demand forecasting, and robot control. Many use cases are already emerging in individual workplaces.
On the other hand, whether this has truly boosted company-wide productivity and profitability is still a work in progress. A survey by S&P Global last year showed that for many companies investing in AI, many initiatives stall before reaching production. On average, 46% of AI projects in companies do not progress from PoC to widespread implementation. McKinsey also notes that many companies remain in the experimental or pilot stage, with only about one-third beginning to scale their AI programs in earnest. Furthermore, only 39% of companies reported that AI has impacted their overall EBIT.
What is important here is the possibility that the issue lies not in the performance of the AI itself, but in how it is used. An IndustryWeek article from March 13, 2026, compares this stagnation to the 'productivity paradox' of the early PC revolution. Initially, simply introducing PCs one by one did not lead to significant productivity gains. The value only truly emerged once they were connected via networks and information began to flow across departments. AI in manufacturing may currently be in that same transitional phase.
If introduced individually, AI will end up as an 'island'
Looking back at the themes covered in this note so far—AI visual inspection, robotics, simulation, skills transfer, and cybersecurity—the discussion points have gradually expanded.
While all are important themes, from another perspective, they are also easily discussed as individual challenges or technologies.
And as long as they remain there, it is difficult to reach the point of significantly changing the productivity of the entire factory. The theme this time is about how to connect those individual points.
However, if each operates independently, the productivity of the entire factory will not change as much as expected. Quality data does not connect to design. Equipment data remains closed within the maintenance department. If production planning and procurement remain siloed, it will only result in an accumulation of local optimizations.
What the IndustryWeek article points out is that the true value of AI in manufacturing lies not in AI closed within individual tasks, but in AI that connects across departments. It is only when generative AI, agentic AI, and machine learning are not separate, but linked across multiple departments and systems, that significant effects are produced. It is not enough for predictive maintenance, quality, logistics, procurement, and design to each have their own separate AI. The key is for AI to function as a 'bridge' rather than an 'island'.
The AMD case study shows that 'connected AI' is better than 'smart AI'
The AMD case study introduced in the IndustryWeek article embodies this image.
According to a public case study by SAP AppHaus last year, at AMD, when a delivery date change occurred for a sales order, up to 14 manual checks were required to trace the cause. By supporting this with generative AI tools on SAP S/4HANA and SAP Business Technology Platform, the time and cost required for root cause analysis were reduced by approximately 90%, with an expected reduction of over 3,100 man-hours per year. What is important here, more than the numbers themselves, is that the AI is not closed within a single department.
It tracks causes and accelerates responses by spanning multiple pieces of information, such as inventory, supply, orders, and customer support. In other words, the AI is working as a 'bridge across departments' rather than a 'tool for streamlining specific tasks.' IndustryWeek further envisions a future where a customer simply takes a photo of a broken part, and the AI links everything from part identification, inventory confirmation, and shipping condition judgment to replenishment arrangements and design improvement proposals. Only by going this far does AI move closer to transforming the entire business flow rather than just local improvements.
The next point for the Japanese manufacturing industry to consider
From here on, these are my own views.
The Japanese manufacturing industry has strengths in improvement at the workplace level. The ability to refine individual processes, reduce waste, and stabilize quality, as seen in 5S and TPM, is extremely high. This is a major strength.
However, precisely because of that strength, AI is easily introduced as 'let's start with this process' or 'let's start with this department,' and as a result, it often remains fragmented. If quality, maintenance, production management, procurement, and design are running on separate systems and data, the effects of introducing AI will remain confined to those islands. The reason it does not reach overall optimization is not because the AI is weak, but because the connection between data and operations is weak.
Therefore, the next question to ask is not just 'which AI tool should we introduce?' Rather, the following three points are important:
Does this AI connect with data from other departments?
Will the effects of this AI ripple through to preceding and succeeding processes?
Is it built on a foundation that can continue to be used even if the model changes in the future?
The IndustryWeek article also lists a high-quality data foundation that can be used company-wide, applications that can incorporate evolving AI models, and a phased progression from simple automation to autonomous collaboration as prerequisites for realizing cross-departmental AI. It is not glamorous, but this order is quite realistic.
Summary
There is no doubt that AI in manufacturing is spreading.
However, the spread of adoption and a significant change in productivity are not the same thing. Many companies are still struggling to transition from experimentation to company-wide value.
The key to crossing that barrier is not to increase the number of individual AIs, but to connect them. Quality information returns to design. Equipment information connects to procurement and production planning. Customer support links with inventory and maintenance. When such AI as a 'bridge' begins to move, the productivity of the entire factory will begin to change.
There is no need to draw up a large-scale plan right now. However, the next time you consider AI, try asking yourself just once, 'Will this AI end as an island, or will it become a bridge?' That perspective alone should significantly change the quality of your investment.
Articles related to this theme
・Automation of major processes from 18% to 50%. Even so, it wasn't technology that made the difference.
・Samsung to make all factories 'AI-Driven Factories' by 2030. How will Agentic AI from Galaxy change manufacturing?
・Siemens introduced AI to design, Bosch to the shop floor | Two entry points for factory AI seen at Hannover 2026
Regarding consultations
I accept consultations as a personal activity, including organizing points before introducing AI for manufacturing, preparation before consulting with AI vendors, and internal study sessions/brainstorming. For details, please see theself-introduction article.
If you use X, I also accept inquiries via DM.
Thank you for reading until the end.
In this note, I organize trends in factory AI, robotics, and smart factories by connecting management, technology, and on-site implementation.
Beyond introducing individual news, I also track which companies enter factory DX from which entry points and where the difficulties of on-site implementation arise.
If you would like to continue reading about this theme in the future, I would appreciate it if you could follow me.
Sources/Reference Links
IndustryWeek | “The Big Productivity Gains Will Come from Cross-Functional AI” |
https://www.industryweek.com/technology-and-iiot/digital-tools/article/55363960/the-big-productivity-gains-will-come-from-cross-functional-ai
S&P Global | “Generative AI experiences rapid adoption, but with mixed outcomes – Highlights from VotE: AI & Machine Learning” |
https://www.spglobal.com/market-intelligence/en/news-insights/research/ai-experiences-rapid-adoption-but-with-mixed-outcomes-highlights-from-vote-ai-machine-learning
McKinsey | “The State of AI: Global Survey 2025” |
https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
SAP AppHaus | “Solving Supply Chain Hurdles with Generative AI on SAP BTP” |
https://apphaus.sap.com/project/solving-supply-chain-hurdles-with-generative-ai-on-sap-btp