ROI Case File No.468 'The Day We Asked Why Five Times'
The Day We Asked Why Five Times
Chapter 1: The Silence of 28%
"Six months after deployment, utilization is stuck at 28%. I don't know what's causing it."
Keiko Noguchi, DX Division Head at TechNova, placed a usage report on the table. A row of bar charts. Utilization had started at 42% in the first month, began declining in month three, and settled into a plateau at 28% six months in.
"What tool was deployed?" I asked.
"Lightblue—a generative AI tool," Noguchi answered. "It handles document creation, email drafting, data analysis support—a general-purpose tool deployable across multiple workflows. We rolled it out to all 200 employees. Training was conducted. But only 56 people are still using it."
"Did you ask the other 144 why?" Claude confirmed.
"We ran a survey," Noguchi answered. "The most common response was 'I don't know what to use it for.' Second: 'I don't know how to use it.' Third: 'I don't have time to use it in my work.'"
"Those three answers are surface-level reasons," Gemini said quietly. "Deeper reasons must lie beneath them."
"What do you mean?" Noguchi leaned forward.
"Why don't they know what to use it for?" Claude said. "That question needs to be asked four more times."
Chapter 2: The Five Layers 5WHYS Digs
"This case calls for 5WHYS."
Claude drew five arrows vertically on the whiteboard: Why ① through Why ⑤.
"5WHYS is a framework that drills from the surface symptom down to root cause by asking 'why' five times," I explained. "A 28% utilization rate is a symptom. Applying measures to a symptom without changing the root cause will not move the number. Five levels down is where the real question lives."
"Let's first measure the cost of the status quo," Gemini said, opening ROI Polygraph and entering the usage logs and training records Noguchi had provided.
The numbers returned.
"The opportunity cost estimate from underutilization is in," Gemini read. "For the 144 non-users, assuming a projected 10 minutes of daily work savings each: at ¥2,800/hour, that's ¥9,300/month per person in opportunity cost. Total for 144 people: ¥1,344,000/month, ¥16,128,000/year lost to non-adoption."
"Additionally," Gemini continued, "inefficiency in one-size-fits-all training cost—the personnel cost for training sessions that produced no results: estimated at ¥756,000/year wasted. Total: approximately ¥17 million in invisible annual cost from this problem."
Noguchi went quiet. "I thought we had taken action by running the training."
"Then let's dig with 5WHYS," I continued.
[Why ① — Why is utilization not rising?]
"The top survey response was 'I don't know what to use it for,'" Claude said. "This is the answer to Why ①. But it describes the symptom, not the cause. Why don't they know what to use it for? On to Why ②."
[Why ② — Why don't they know what to use it for?]
"What did the training cover?" I asked Noguchi.
"How to operate the tool and how to write basic prompts," she answered. "Two hours of classroom-style lecture, same content for everyone."
"That's it," Claude said. "The training ended with a tool explanation—without connecting it to actual work. Telling a sales rep 'this tool can write documents' means nothing if they can't see 'how do I use this for my own proposals?' Why ② is that training never connected to work."
[Why ③ — Why didn't training connect to work?]
"Which team designed the training?" Gemini confirmed.
"The DX division. We designed it ourselves," Noguchi answered. "I thought training everyone on the same content was more efficient."
"Why ③ is visible," Claude continued. "The DX division that designed the training didn't understand each department's work. Sales, accounting, manufacturing—without knowing what the pain points were in each, they built one-size-fits-all content. A gap in operational context existed between the designers and the participants."
[Why ④ — Why did the context gap form?]
"Did you conduct per-department interviews before designing the training?" I asked.
Noguchi paused. "No. There wasn't time, and—honestly—we didn't have the capacity."
"That's Why ④," Claude said. "The DX division was under-resourced. Trying to reach 200 employees with a small team meant per-department design was impossible. Resource constraints created the one-size-fits-all design."
[Why ⑤ — Why is the DX division under-resourced?]
"How many people are in the DX division?" I confirmed.
"Three," Noguchi answered. "Recruiting people with AI knowledge has been difficult. The market for AI talent is intensely competitive, and our compensation conditions make it hard to attract candidates."
"The root cause at Why ⑤ is in," Claude said quietly. "Difficulty recruiting AI talent creates chronic under-resourcing in the DX division, forcing the compromise of one-size-fits-all training. The root cause of 28% utilization was not the tool, not the employees—it was a structural organizational problem."
Noguchi exhaled deeply. "Five levels down and I ended up somewhere completely different."
Chapter 3: Three Prescriptions for the Root Cause
"Now that the root cause is clear, let's design the prescriptions," I said, stepping to the whiteboard. "But AI talent recruitment difficulty—that root cause cannot be changed quickly. That's exactly why we need a design that functions despite the recruitment constraint."
"Three prescriptions," Claude said.
[Prescription ①: Develop Internal Evangelists by Department]
"The problem is that the DX division is trying to serve every department—that's what stretches the resource," Claude continued. "Identify one employee per department to act as an evangelist—someone who learns the AI tool ahead of others and leads adoption within their department. The DX division focuses entirely on developing evangelists. Not trying to reach 200 directly—focusing on developing 10 evangelists. That is the best strategy within resource constraints."
"How should we select evangelist candidates?" Noguchi asked.
"From within the 56 who are already using it," Gemini answered. "Among current users, there will be someone experimenting autonomously and enthusiastically. That person is best positioned to show others. They're already doing it."
[Prescription ②: Design Department-Specific Use Cases in Advance]
"Before training, determine one concrete use case per department," I continued. "Sales: first draft of proposal documents. Accounting: monthly report summarization. Manufacturing: revision of operation manuals. Narrowing to one is essential. One success experience opens the door to the next."
"Let's reference same-scale company data through Strategic ROI Intelligence," Gemini proposed. "It surfaces which use cases by department have shown the highest impact—choosing from proven options is more persuasive to evangelists than designing from scratch."
[Prescription ③: Redesign Training in Three Levels]
"Restructure the current one-size-fits-all training into three levels," Claude said. "Beginner: starts from 'what is AI' and runs through one complete use case experience. Intermediate: can write prompts applicable to their own work. Advanced: can spread adoption within their team. The same two hours of training time—split by audience—produces a completely different absorption rate."
"Let's project the post-improvement numbers with ROI Proposal Generator," Gemini proposed.
The projection assumed a target utilization rate of 65% after evangelist deployment.
- New active users: 200 × 65% − 56 = 74 additional active users
- Monthly improvement: 74 × 10min × ¥2,800 ÷ 60 × 20 days = ¥692,000/month
- Training efficiency gain: ¥756,000/year saved
- Total annual improvement: approx. ¥9.05M
- Intervention investment (evangelist training, use case design): est. ¥1.2M
- Payback period: approx. 1.6 months
"Payback in under two months," Noguchi said.
"The root cause is AI talent recruitment difficulty," Claude replied, "but by developing evangelists from internal talent, utilization can be raised without external hiring. If external hiring is difficult, develop what's inside. That is the realistic answer to Why ⑤."
Chapter 4: What the Fifth Why Changed
Noguchi stood and looked at the Why ① through Why ⑤ arrows on the whiteboard.
"At first I thought the problem was that employees weren't used to AI. That training was insufficient. But five levels down: it was training design, caused by resource shortage, rooted in hiring difficulty. The problem was in an entirely different location."
"That is the essence of 5WHYS," I replied. "Applying a measure to the visible problem causes the problem to return in a different shape. Digging to the root changes what the measure needs to be. Improving one-size-fits-all training would have triggered the same outcome again. Knowing the root creates a path even within constraints."
"Starting next week, I'll interview candidates in each department," Noguchi said. "The first task is selecting ten evangelists from the 56."
Outside the window, office building lights were beginning to switch on one by one.
Five months later, a report arrived from Noguchi.
Three months after ten evangelists were placed in each department, company-wide utilization rose from 28% to 61%. Evangelists demonstrating use cases within their departments caused "I don't know what to use it for" responses to nearly disappear. The department that showed the most impact was Sales: average proposal writing time shortened by 40 minutes.
Training was redesigned into beginner, intermediate, and advanced levels. Participant satisfaction rose 31 points from the previous session. Noguchi and the three DX division members were freed from direct engagement with all 200 employees and could now focus on designing the next wave of AI utilization initiatives.
The final line of Noguchi's report read: "Asking why five times changed where the measure landed. Rather than acting directly on the utilization number, we changed the structure producing that number. When the structure changed, the number followed naturally."
The day we asked why five times was the day change began.
"Problems cannot be solved where they appear. The visible problem is a symptom; the root cause sleeps five layers below. What 5WHYS demands is the patience to keep asking. The first why teaches you the symptom. The second shows the entrance to the cause. By the third, structure starts to appear. The fourth reveals the constraint. The fifth is where root cause shows its face. Only the place reached by asking all five is where a real prescription can be written."
Related Files
Tools Used
- ROI Polygraph — AI tool underutilization opportunity cost and training inefficiency visualization
- Strategic ROI Intelligence — Same-scale company per-department AI use case reference
- ROI Proposal Generator — Evangelist development initiative payback simulation
