ROI Case File No.499 'The Contract You Were Looking For, in Thirty Seconds'
The Contract You Were Looking For, in Thirty Seconds
Chapter One: The File Server Ends Next Year
"The file server service we use ends support next March."
Wataru Usami, IT department director at Globex Pharmaceuticals, showed a notice from the vendor. The file server they had used for over 10 years would cease operation at the end of next fiscal year. "We need to decide a migration destination, but migrating with the current folder structure carries the underlying problem forward."
"What's the current problem?" Claude confirmed.
"Capacity is tight," Usami answered. "The same documents are stored in multiple locations, and no one has authority to organize. Many documents can't be found through search, and finding them takes time every time. Furthermore, there are multiple CSV connections and IP-address-dependent system integrations, and the impact of migration can't be fully assessed."
"With pharmaceutical manufacturing, document management has regulatory implications," Gemini asked.
"Large volumes of GMP-compliant documents," Usami answered. "Quality management department business records, manufacturing records, change management records—these have storage obligations under regulations. Searchability of records directly affects business efficiency."
"Whether to integrate work files (in-progress) with formal documents hasn't been decided either," Usami continued. "Separating makes search two-stage. Integrating loosens security. We have no axis for judgment on which is correct."
"We need to first see where waste is," I said. "LEAN's turn."
Chapter Two: LEAN Asks About Waste, Overburden, and Unevenness
"This case requires LEAN."
Claude wrote four letters on the whiteboard: L, E, A, N.
"LEAN is a way of thinking about business improvement born from the Toyota Production System, a framework for eliminating muda (waste), muri (overburden), and mura (unevenness) and flowing value," I explained. "It's a manufacturing field methodology, but it can be applied to information processing operations. Document management is an area where waste (search time, duplicate storage, storage of unnecessary documents), overburden (impossible organization rules, complex permission design), and unevenness (different rules per department, search ability differences between people) easily occur. Decomposing through LEAN clarifies system requirements."
"Let's first measure current costs," Gemini said, opening ROI Polygraph. She entered the data from Usami.
"Monthly document management costs are out," Gemini read. "Time spent on document search averages 6 hours per person per month. For 200 employees, 1,200 hours per month. At average 3,500 yen per hour, equals 4.2 million yen monthly. Duplicate storage management workload: 70 hours, equaling 245,000 yen. GMP audit response workload: 100 hours, equaling 420,000 yen. Server capacity expansion and organization response from tightness: 40 hours, equaling 160,000 yen. CSV integration incident response: 20 hours, equaling 80,000 yen. Total: 5,105,000 yen monthly. Annualized: approximately 61.3 million yen."
Usami studied the figures. "Search time alone is 4.2 million yen monthly. This wasn't visible."
"Now let's design with LEAN," I continued.
[Eliminating Waste—Structural Resolution of Duplication and Search Cost]
"First, we decompose document waste," Claude said. "Duplicate storage, dead storage of past documents, search difficulty—three types of waste. Duplication resolves through unified version management in the document management system. Dead storage resolves by setting retention period rules and auto-archiving expired documents. Search difficulty resolves through full-text AI search and automatic tag assignment."
"AI search?" Usami confirmed.
"Search that understands document content," Gemini answered. "Find through context even without exact keyword matches. For example, even with vague searches like 'last year's contract with company XX,' it hits. Even in areas with many specialized terms like GMP documents, search by document meaning is possible."
[Resolving Overburden—Integrating Work Files and Formal Documents]
"Overburden arises from impossible separation design," I continued. "Completely separating work files and formal documents creates the need to traverse two locations to find in-progress documents. Make it an integrated system, with design managing by document status (draft, in review, approved, formal). Search completes in one screen."
"How is security ensured?" Usami confirmed.
"Set access permissions by status," Claude responded. "Formal versions have wide viewing scope, drafts limited to creator and approval route. Permission management within integration ensures security equivalent to separation."
[Organizing Unevenness—Unifying Rules Between Departments]
"Unevenness arises from different folder configurations and naming rules per department," Gemini continued. "Conduct company-wide unified metadata design. Document type, related departments, related projects, retention period—mandate these as tags. Search not dependent on folder structure is realized through metadata."
"Field rule changes will face resistance," Usami worried.
"We reduce burden through automatic tag assignment during migration," I answered. "AI analyzes existing documents and auto-assigns metadata. Not fully automatic, but 80 percent automatic with 20 percent human-checked design. Manual full-volume tag assignment isn't realistic."
[Calculating Investment Recovery]
"Let's run the numbers with ROI Proposal Generator," Gemini suggested.
- Initial cost: Document management system implementation, AI full-text search infrastructure, migration work, automatic tag assignment processing, GMP compliance configuration, all-employee training. Total: 11.8 million yen
- Monthly cost: Document management system, AI search infrastructure combined: 360,000 yen
- Monthly reduction: Search time reduction = 2.94 million yen (assuming 70% reduction), duplicate management workload reduction = 170,000 yen, GMP audit response reduction = 250,000 yen, server expansion response reduction = 160,000 yen, integration incident response reduction = 50,000 yen. Total: 3.57 million yen monthly
- Net monthly reduction: 3,570,000 − 360,000 = 3,210,000 yen
- Payback period: 11,800,000 ÷ 3,210,000 = approximately 3.7 months
"Payback within four months," Gemini summarized. "The reason for short-term payback is the large search time reduction effect. At 200-person scale, even a few hours per person reduction accumulates very largely."
Usami confirmed the figures. "I was viewing this as file server migration cost. Organizing through LEAN, the visibility changes to investment in business improvement."
"Resolving waste, overburden, and unevenness lets server migration proceed as a secondary effect," I responded.
Chapter Three: Turning Migration into an Opportunity for Organization
"Let me organize the approach," I said, standing at the whiteboard.
"Months 1–2—Document management system selection, conduct PoC. Month 3—Metadata design, retention period rule formulation, GMP compliance requirement confirmation. Months 4–5—Automatic tag assignment processing of existing documents, human correction. Month 6—AI full-text search accuracy adjustment, employee training. Month 7—Begin parallel operation. Month 8—Old server stop, new system standalone operation. Month 9 onward—Operation optimization and continuous improvement."
"What's the business impact during migration?" Usami confirmed.
"We set a one-month parallel operation period," Claude answered. "A state where both old server and new system are accessible. If documents have incomplete migration, they can be retrieved from the old server. Complete cutover happens after parallel operation stabilizes."
Usami took notes and said, "I was being chased by the server end deadline. Through LEAN, a perspective of changing the deadline into business improvement emerged."
Chapter Four: The Day the Contract You Were Looking For Was Found in Thirty Seconds
Eleven months later, a report arrived from Usami.
Document search time decreased by 73 percent compared to before, three months after launch. AI full-text search allowed reaching the target document even with vague memory-based searches. "Even with memory-based searches like 'a contract with the XX supplier last year or the year before,' it comes out in 30 seconds," Usami wrote.
GMP audit response workload also dramatically decreased. Documents requested in audit demands could be filtered by metadata and instantly extracted. "Previously we spent three days gathering documents before audits. Now during the audit meeting, we can receive the request and produce them on the spot," the report noted.
The biggest change appeared in the quality of document creation processes. Because AI search instantly finds past similar documents, the need to write new documents from scratch disappeared. Operations of polishing past excellent documents as a base took root, and document quality became uniform across departments. "Good documents written by veterans became referenceable assets for younger staff too," the report noted.
The integrated design of work files and formal documents was also accepted in the field. Status management allowed in-progress documents and formal documents to be searched from the same screen. "The hassle of going back and forth between two locations disappeared," many comments came in user surveys.
As a secondary effect, visualization of internal knowledge progressed. Analyzing AI search query logs revealed what kind of information employees seek. For documents frequently searched but hard to find, dedicated navigation was added. "Search logs became a guide for internal information maintenance," Usami wrote.
Server migration completed on schedule in month 8. CSV integrations and IP-address-dependent peripheral systems were also organized through migration. While reviewing integration destinations, two systems were identified as unnecessary and decommissioned. Integration incident response workload also naturally decreased. "I was being chased by the server end next March, but in retrospect, it was an opportunity for overall optimization," the report noted.
The end of Usami's report read: "Move because waste exists, not because the deadline arrives. Through LEAN's lens, server end was not the end—it was the entrance to organization."
The state where the contract being searched for came out in 30 seconds had become the norm.
"Document management easily becomes a nest of waste. Duplicate storage, dead-stored past documents, unfindable searches—seemingly small waste becomes 4.2 million yen monthly at 200-person scale. What LEAN asks is the three-layer decomposition of waste, overburden, and unevenness. Waste resolves through structure, overburden through design, unevenness through rules. Whether the server end deadline can be transformed into an opportunity for organization is a matter of perspective. The morning the contract being searched for came out in 30 seconds, what disappeared wasn't time—it was the act of searching itself."
Related Files
Tools Used
- ROI Polygraph — Visualizing document search workload, duplicate management, and audit response cost
- ROI Proposal Generator — Investment recovery simulation for document management system implementation
