Executive Management Leadership Ver07. “The Day I Won the World's First Fully Productized Digital Solution for a Foundry” – A PMO Story of Stepping Forward into a New Field
- shigenoritanaka3
- 5 日前
- 読了時間: 4分
Apr 01, 2026
Thank you again for reading.
This time, I would like to share the story of how I won the world's first order for a fully productized digital solution for a sand-mold foundry - an achievement realized through a PMO project at a foreign‑affiliated manufacturer.
For many years, a foundry customer repeatedly told me:
“We have PLCs, so the data exists. But we still have no visibility into what is happening in the plant. We need real‑time visualization.”
Our European headquarters had been developing such a system, but the product launch was delayed, and customer expectations were fading.
Then one day, the head of digital sales from HQ suddenly visited Japan and informed us that a fully productized digital solution - combining visualization with AI-based defect-reduction recommendations - was finally complete.
Prototypes had already been sold in South Africa and Europe, but this project became the first official sale of the full product version including AI.
In 2019, I proposed the solution to the customer. They showed strong interest in visualization but did not see the need for AI.
At that moment, I made a decision:
“How will you recover the digital investment? The answer should be defect reduction.”
I built an ROI (return on investment) model, and with government subsidies, the customer decided to invest. I then managed to secure the contract.
** For the benefits and limitations of factory digitalization, see: Factory Digitalization_Ver01. _ "The Benefits and Limitations of Digital Transformation in Sand Casting Foundries" - What Should Be Adopted Together?
Strong Internal Opposition
Within the Japan subsidiary, everyone opposed taking the order.
“Tanaka‑san, if we accept this, who will take responsibility?” “Is the technology even complete? You'll get stuck in a swamp.”
I responded:
“Even if the data upload fails, it won't affect the customer's production. I will take responsibility. No one else will be burdened.”
That was the moment I committed to stepping forward.
Stepping Alone into a Field Outside My Expertise
At the time, I was both the Japan Country Manager and Controller - already handling two mission-critical roles. I had no background in foundry processes or digital IoT.
Even so, I entered the plant as PMO, learned the processes, and - thanks to the customer's engineer who mapped the entire LAN and CC‑Link network - gained a full understanding of the data flow.
With that support, I defined system requirements, organized thousands of data points (names, scaling factors, units, PLC addresses, data sources), translated them into English, and collaborated with HQ to visualize all data across equipment from multiple manufacturers.
Driving Everything from AI Model Building to User Training
Working closely with HQ's partner AI company in South Africa, I led:
Data ingestion adjustments
Batch-level data integration starting from pouring
Linking inspection results to build AI learning models
Reviewing AI recommendations
Releasing the digital site to the customer
Dashboard development
Training managers and operators
Of course, none of these would have been possible without the customer's highly skilled engineer who supported the network and data logging aspects.
The Result: Full Visualization and AI Validation
After more than a year, with the cooperation of HQ and the customer, full plant visualization was completed.
In 2021 - despite the COVID pandemic - digital engineers, foundry application engineers, and AI engineers from South Africa came to Japan to conduct AI trial runs and validation.
The result:
Approximately 40% defect-reduction potential was demonstrated.
But the Real Challenge Began After AI Validation
AI validation is not the finish line. The real challenge is:
Embedding AI into daily operations to continuously reduce defects.
I monitored inspection results daily through the web interface.
Many records showed inspection progress over 100% (like over 200%!) → Incorrect manufacturing date entries
Defect rates varied significantly by inspector → Inspector-level variation
This meant that distorted inspection data severely degrades AI accuracy.
** For why quality data “lies,” see: Factory Digitalization_Ver03. _ “When Quality Data Lies – How Inadequate KPIs Create the Wrong Behavior in Manufacturing”
AI is ultimately mathematics. If the learning model is flawed, no algorithm can deliver accuracy.
That is why I continued improvement efforts for six years - introducing mechanisms to prevent input errors, reducing inspector variation, and providing user training.
For new products (as I will write in another article), AI operation is impossible. So, I repeatedly discussed realistic improvement methods with managers and operators and continued operational refinement.
** For more details, see: Factory Digitalization Ver06. “AI Is Weak with New Products.”- So What Should Manufacturers Do?
Through these efforts, trust was built, and the digital solution contributed to equipment renewal and aftermarket business. As a result, the customer achieved record-high profits last fiscal year.
“This Was Not a Digital Project — It Was a Leadership Project”
Looking back, this was not about technology.
It was about stepping forward when no one else would, embracing uncertainty, taking on the PMO role, and connecting the plant with HQ to move the project forward.
In an era where equipment is commoditized, providing digital services added meaningful value to traditional equipment and aftermarket business.
More details will be shared in a future article.
Consultation
If you face challenges in PMO or digital/AI implementation— especially in areas where “no one wants to step forward”— I would be glad to support quietly and professionally.
Contact: info@metricjapan.com
TAGs
コメント