ROI Case File No.578 'The Burden Was Visible, but We Couldn't See the Four Distances'
The burden was visible, but we couldn't see the four distances
Chapter 1: Checking is hard work—but what is making it so?
"We want to analyze our network-camera footage with AI, to automatically detect dangerous or negligent acts."
Tō Mamiya, the manufacturing-department manager at Globex Corporation, described the situation. "We take reports from employees through an internal electronic suggestion box. But among the reports of violence or negligence, many differ from the facts. So, to confirm the risk, we always watch the footage. Because the reports don't specify a time, we check long footage on fast-forward. This is a heavy burden."
"How are you trying to reduce that burden?" Claude asked.
"By automating the footage checking with AI," Mamiya answered. "Have AI take over the fast-forward labor. But why the footage check is always required in the first place, why we can't watch the whole floor—I haven't fully grasped the background of the burden."
"In the background of that burden, do you see the distances of culture, administration, geography, and economics?" I confirmed.
"…I don't," Mamiya answered. "I've only been looking at the burden of footage checking. Why the check becomes mandatory, why monitoring is difficult. I've never separated out the background distances."
"The burden is visible, but unless you can see the four distances, you can't compose an effective hand," I replied. "Let's break it down with CAGE."
Chapter 2: What CAGE asks—grasp the background across four distances
"This case calls for CAGE."
Claude wrote on the whiteboard: "Cultural, Administrative, Geographic, Economic."
"CAGE—the four distances of cultural, administrative, geographic, and economic—is a framework that grasps a problem's background from four distances," I explained. "The crux is not looking only at the surface burden. Treat 'footage checking is hard' and you drop the background. Why is the check mandatory (cultural), how about the administration, why is monitoring difficult (geographic), what does reduction produce (economic). Because you grasp the background across four distances, you can compose an effective hand. It's a tool for separating out the background."
"First, let's measure the current cost," Gemini said, opening ROI Polygraph. He entered the data Mamiya had provided.
"The monthly cost is out," Gemini read off. "Fast-forward footage-checking labor averages 185 hours a month; at ¥3,800 an hour, that's ¥703,000 a month. The burden of fact-checking anonymous reports averages ¥380,000 a month. Oversight risk from the difficulty of real-time monitoring averages ¥360,000 a month. Accident and loss risk from leaving dangerous and negligent acts unaddressed averages ¥400,000 a month. Stalling where the system exists but improvement measures are lacking averages ¥300,000 a month. The total is ¥2,143,000 a month—about ¥25.72 million a year."
Mamiya stared at the figures. "I'd been watching only the footage-checking labor. Once you add the cost of dropping the background distances, it comes to this much."
"Then let's design it with CAGE," I continued.
[Cultural—grasp the background that makes the check mandatory]
"First, the cultural distance," Claude said. "Trust among employees is thin, and anonymous reports are heavily used. So reports that differ from the facts get mixed in, and a footage check for confirmation becomes mandatory. This culture is in the background of the burden. Grasp it first."
[Administrative—grasp the gap in the reporting system]
"Next, the administrative distance," Gemini continued. "The reporting system is in place, but concrete improvement measures are lacking. Even with a system, no technology to lighten the manager's burden is built in. Grasp the gap in the administration."
[Geographic—grasp the difficulty of monitoring]
"After administration, the geographic distance," I continued. "The manufacturing floor spans a wide area, and real-time monitoring is difficult. The human eye can't watch it all. That's why automatic detection through AI analysis works. Fill the geographic distance with technology."
[Economic—grasp the effect that reduction produces]
"Last, the economic distance," Claude continued. "A reduction in labor time and greater efficiency translate directly into cost reduction. AI introduction requires an initial investment, but seen over the long term the ROI is high. Having grasped the four distances, land it on the economic effect."
[Estimating the payback]
"Let's run the numbers with the ROI Proposal Generator," Gemini proposed.
- Initial cost: CAGE analysis, AI-footage-analysis introduction, auto-detection/alert features, footage search and management, and operational design—¥5.5 million total
- Monthly cost: analysis-system operation plus ongoing maintenance—¥240,000
- Monthly savings: automation of footage checking = ¥560,000 (assuming a 70% reduction); efficiency of fact-checking = ¥340,000; oversight prevention through real-time detection = ¥300,000; loss avoidance through accident prevention = ¥300,000; ¥1,500,000 a month total
- Net monthly savings: ¥1,500,000 − ¥240,000 = ¥1,260,000 a month
- Payback period: ¥5.5 million ÷ ¥1,260,000 = about 4.4 months
"A little over four months to recover," Gemini summarized. "What works is not looking only at the burden of footage checking but grasping the background across four distances. Merely have AI take over the burden, and the background—anonymous reports, a sprawling floor—remains. Because you grasp it across four, automatic detection works on the background. The investment doesn't swing at empty air."
Mamiya said as he checked the numbers. "I thought reducing the fast-forward labor would settle it. Unless you see the four distances, the background remains."
"CAGE is a tool for grasping the background across four distances," I replied.
Chapter 3: An implementation plan that grasps across four distances
"Let me lay out the approach," I said, standing before the whiteboard.
"Month one—background analysis across the four distances and definition of the detection targets. Month two—requirements definition for AI footage analysis and design of automatic detection. Months three and four—introduce the analysis software and implement alert notification. Month five—strengthen the footage-data search and management features. Month six—trial run and effect verification. Month seven onward—improve detection accuracy and roll out across the whole floor."
"If we automate the footage checking with AI, the burden decreases—right?" Mamiya confirmed.
"Automation alone isn't enough," Claude replied. "Unless you grasp the background that generates the burden—a culture that heavily uses anonymous reports, a wide, hard-to-monitor geography—the checking itself won't decrease. Grasp the four distances with CAGE, and you make automatic detection work on the background and can objectively evaluate the facts. Grasping the background comes before automation."
Mamiya said as he took notes. "Before automating the footage check, grasp the background across the four distances. I can see the order now."
Chapter 4: The day, grasping the distances, the burden decreased
Ten months later, a report arrived from Mamiya.
Footage checking lightened greatly after the AI analysis was introduced. "Where we'd watched long footage on fast-forward, AI began to detect the relevant acts automatically. The time taken up in checking vanished," Mamiya wrote.
Fact-checking became objective too. Automatic detection backed the truth of anonymous reports with footage. "We were no longer jerked around by reports that differed from the facts. We could evaluate acts objectively," the report said.
The biggest change showed in how the burden was grasped. From seeing only the labor of footage checking, they shifted to grasping the background across four distances. "I'd been stuck at 'reduce the fast-forward labor.' Once I separated the background by cultural, administrative, geographic, and economic, I could even see why the check becomes mandatory, and could work from the root," Mamiya wrote.
Oversight decreased too. Real-time automatic detection filled in the monitoring of the wide floor. "The floor the human eye couldn't watch could now be captured by AI. Oversight of dangerous and negligent acts decreased," the report said.
As a side effect, the way problems were viewed changed. Rather than the surface burden, the habit of grasping the background distances took root. "I stopped ending at 'reduce the labor.' I began to think, across the four distances, about what, and why, generates that burden," Mamiya wrote.
At the end of Mamiya's report was written: "I thought the manufacturing-floor trouble was the burden of footage checking. But the real problem was that the burden was visible while I couldn't see the four distances. The moment I separated the background into four with CAGE, the effective hand came into view. Before automating the footage check, grasping the background across four distances came first."
On the day a company where the burden was visible but that couldn't see the four distances became a company that could grasp the background across four distances, the manufacturing-floor efficiency had changed from automating footage checking into a design that grasps the background across four distances—cultural, administrative, geographic, and economic.
"A consultation about the manufacturing floor usually arrives in the form of 'we want to reduce the burden of checking.' But before automating, there's something to ask. Is the background that generates that burden visible? What CAGE asks is the four distances of cultural, administrative, geographic, and economic. Grasp the background across four and automatic detection works on the root. On the day a company where the burden was visible but that couldn't see the four distances could grasp the background, what changed was not the speed of footage processing but the very perspective of grasping the background across four distances."
Related Files
Tools Used
- ROI Polygraph — visualizing footage-checking labor, fact-checking burden, and accident risk from unaddressed acts
- ROI Proposal Generator — a payback simulation for manufacturing-floor efficiency, starting from a background analysis across the four distances
