ROI Case File No.538: When That Person Was Out, the Trucks Had Nowhere to Go
When That Person Was Out, the Trucks Had Nowhere to Go
Chapter 1: Without the Veteran, Dispatch Stops
"Our dispatch work is completely individual-dependent. When that one person is out, the trucks have nowhere to go."
Takashi Tomabechi, operations management director at TransLogix, spoke while showing route maps. A logistics company specializing in frozen-food transport. "We run long-haul between Tokyo and Osaka, and local delivery from bases in Aichi, Mie, Osaka, and the Kanto region. The person at each branch builds the dispatch by hand."
"How deep is the individual dependence?" Claude asked.
"Only a veteran can build it," Tomabechi answered. "When that person is on leave or out sick, there's no substitute. Even emergency dispatch changes take time. We can't compare multiple dispatch patterns on the spot, either. We can't respond flexibly to customer order changes."
"Have you tried a system before?" I confirmed.
"We tried and shelved it," Tomabechi answered. "We trialed AI auto-dispatch, but it couldn't handle our operating conditions—accuracy was around 50–60 points. We still check partner-company vehicle locations by phone. The issues are mountainous. But which one to tackle to make the investment pay—I can't judge."
"If the issues are many, you need to judge the investment by reach, impact, confidence, and effort," I responded. "Let's break it down with RICE."
Chapter 2: RICE Asks Reach, Impact, Confidence, and Effort
"This case needs RICE."
Claude wrote "R・I・C・E" on the whiteboard.
"RICE is a framework that scores measures on four axes—Reach, Impact, Confidence, Effort—and decides priority," I explained. "When the issues are many, choosing by impact alone tempts you to leap at big jobs you're not confident in. RICE judges the move worth the investment by multiplying four things—how widely it works, how large the impact, how much you can be confident of success, and how much effort it takes."
"Let's measure the current cost first," Gemini said, opening ROI Polygraph and entering the data Tomabechi had provided.
"The monthly dispatch-work cost is in," Gemini read out. "Manual dispatch work averages 240 hours a month; at ¥4,800 an hour, that's ¥1,152,000 a month. Response delay and substitute-securing when the person is out averages ¥800,000 a month. Inefficient routing and cost overruns from being unable to compare dispatch patterns averages ¥900,000 a month. Delay loss in emergency order-change response averages ¥550,000 a month. The inefficiency of phone-checking partner vehicles averages ¥350,000 a month. The expected value of key-person risk in dispatch know-how averages ¥600,000 a month. The total is ¥4,352,000 a month—roughly ¥52.22 million a year."
Tomabechi stared at the figures. "I thought it was only the labor of dispatch. Add the cost overruns from inefficient routing and the key-person risk, and the scale is different."
"Then let's design it with RICE," I continued.
[Reach—How Widely Does It Work]
"First, we score Reach," Claude said. "It works on a fleet of several hundred vehicles, including all branches and partner companies. Because dispatch is core work common to every base, the reach is extremely wide. Not a partial improvement but an impact on the whole company's dispatch. Reach is the measure of 'how many it works on.'"
[Impact—How Large the Effect]
"Next, Impact," Gemini continued. "Dispatch efficiency, swift emergency response, flexible handling of order changes, and resolving individual dependence. The effect of dispatch not stopping even when the person is out is large. Impact is the measure of 'how much it works per case.'"
[Confidence—How Sure of Success]
"Confidence, we view carefully," I continued. "There's the fact that the past trial scored 50–60 points. So we can't set it high unconditionally. Still, with customization that meets the operating conditions and selection of a reliable vendor, confidence can be raised. Scoring the past failure honestly as a fact that lowers confidence is RICE's integrity."
[Effort—How Much Effort It Takes]
"Finally, Effort," Claude continued. "From selection through customization, deployment, and operation, it takes considerable time and cost. The effort isn't small. But because Reach is wide and Impact is large, multiply the four axes and the numbers show the effort is worth investing. Effort is the denominator that discounts 'how heavy it is.'"
[Estimating the Payback]
"Let's run the numbers with ROI Proposal Generator," Gemini proposed.
- Initial cost: building the AI auto-dispatch system, operating-condition customization, vehicle location-data linkage, PoC environment build, and training for all branches—¥9.4 million total
- Monthly cost: system operation and ongoing model updates combined—¥380,000 a month
- Monthly savings: dispatch-work hours cut = ¥810,000 a month (assuming 70% reduction); out-of-office response cost cut = ¥560,000 a month; inefficient routing cut = ¥700,000 a month; emergency response accelerated = ¥400,000 a month—¥2,470,000 a month total
- Net monthly savings: ¥2,470,000 − ¥380,000 = ¥2,090,000 a month
- Payback period: ¥9.4 million ÷ ¥2,090,000 = about 4.5 months
"Payback in four and a half months," Gemini summarized. "What matters is reflecting the past 50–60-point shelving honestly in confidence. That's exactly why we design on the premise of operating-condition customization. Multiply the four axes with RICE and, because Reach is wide and Impact is large, it comes out worth the investment even viewing confidence carefully."
Tomabechi said as he checked the figures, "Having failed once, I'd been hesitating. If it scores confidence low and still comes out worth it in numbers, I can step forward."
"RICE is a tool for judging investment by scoring even past failures honestly," I responded.
Chapter 3: A Rollout Plan Starting from a PoC
"Let me lay out the approach," I said, standing at the whiteboard.
"Months one and two—inventory the dispatch work, identify the operating conditions the past trial couldn't handle, and select a vendor. Months three and four—customize to the operating conditions and design vehicle location-data linkage. Month five—a PoC at one branch and accuracy verification. Month six—evaluate the PoC results and re-score confidence. Month seven—after confirming the effect, plan rollout to all branches. Month eight onward—roll out sequentially, expand partner-vehicle linkage, and keep improving the model."
"What if the accuracy isn't there again?" Tomabechi confirmed.
"That's why we narrow the PoC to one branch," Claude responded. "The past failure may have been caused by deploying widely without meeting the operating conditions. We first verify at one base whether the conditions can be met, and re-score confidence on real data. If it isn't there, we revisit the customization. Far smaller loss than failing after expanding company-wide. It's a structure where RICE's Confidence is measured at the PoC."
Tomabechi said as he took notes, "I'd stopped out of fear of failure. If we can advance while re-measuring confidence, we can move."
Chapter 4: The Day the Trucks Move No Matter Who Is Out
Ten months later, a report arrived from Tomabechi.
After the PoC led to full rollout, manual dispatch work was down 70% versus before. "AI now builds most of the dispatch. The person only handles exceptional adjustments. The veteran's time returned to work that requires judgment," Tomabechi wrote.
The out-of-office problem was resolved too. Individual dependence dissolved, and dispatch no longer stopped no matter who was out. "The state where the trucks had nowhere to go when that person was out is gone. We're chased less by securing substitutes," the report said.
The biggest change showed up in emergency response speed. They could now compare multiple dispatch patterns on the spot. "We can respond to a customer's order change with several options on the spot. Before, rebuilding took time," Tomabechi wrote.
The PoC judgment worked too. Because they verified accuracy at one branch before expanding, they avoided a repeat of the past. "At the PoC we confirmed in numbers that 'this time we can meet the operating conditions.' So the budget for company-wide rollout passed. Precisely because we'd failed once, the order of verify-then-expand worked," the report said.
As a side effect, grasping partner vehicles advanced. Location info that had been phone-checked became visible in the system. "We know where partner vehicles are without phoning. Dispatch accuracy rose," Tomabechi wrote.
At the end of Tomabechi's report, he had written this: "There were too many issues, and we'd failed once. So we couldn't move. We scored the four axes with RICE, viewed confidence honestly low, and advanced while re-measuring at a PoC. Build the failure into the scoring without hiding it, and the next investment can be judged."
The day a company where the trucks had nowhere to go when that person was out became one where dispatch moves no matter who is out, dispatch had turned from inside a craftsman's head into a mechanism you can rebuild anytime, the report read.
"A once-failed measure carries a second hesitation. The issues are mountainous, yet the past failure binds the investment decision. What RICE asks is reach, impact, confidence, and effort. How widely it works, how large the impact, how sure of success, how much effort. Score the past failure honestly as a fact that lowers confidence, and re-measure at a PoC. When a company where the trucks stopped when that person was out got the day dispatch moves no matter who is out, what changed was not the dispatch system but the very judgment that builds failure into the scoring and judges the investment."
Related Files
Tools Used
- ROI Polygraph — Visualizing dispatch-work hours, inefficient routing, and key-person risk
- ROI Proposal Generator — Payback simulation for AI auto-dispatch deployment starting from a PoC
