ROI Case File No.458 'The Tender With No Question'
The Tender With No Question
Chapter 1: Three Departments, Three Different Answers
"I sent the same question to three departments and got three different answers."
Shinichi Kawaguchi, Business Reform Manager at Global Meditech, spread three printed emails across the table as he spoke. The same question — "What documents are required for inpatient admission?" — had been sent to Medical Records, the Ward, and Outpatient. The three responses sat side by side. The answers differed subtly, and substantially.
"Knowledge silos in internal inquiry response are serious. There are manuals, but they're hard to follow, and the right department to ask varies. When the person in charge is away, responses stop entirely. We're looking at chatbot implementation to improve this situation."
"Before the chatbot—" I confirmed, "do you have a handle on inquiry volume and the labor cost of responding?"
Kawaguchi paused briefly. "We don't have precise numbers. But based on conversations with staff in the three main departments, roughly thirty inquiries per day across all three. At an average of ten minutes each, that works out to about five hours per day spent on inquiry response across the three departments."
"Monthly," Gemini calculated, "twenty working days — a hundred hours. For inquiry response alone, across three departments."
"What percentage of queries do you think a chatbot could handle?" Claude asked.
Kawaguchi considered. "Recurring, routine questions are probably sixty to seventy percent of total volume. Admission paperwork, procedure flows, directing to the right department — those kinds of questions come in nearly every day in similar form."
"So," I summarized, "if the sixty percent automatable by chatbot is handled, sixty hours per month returns to medical staff. That number is the starting point for today's request."
Chapter 2: RFP Creates the Question
"This case calls for an RFP."
Claude wrote on the whiteboard in large letters: RFP — Request for Proposal.
"An RFP is the document submitted when requesting proposals from vendors," I explained. "But the real value of an RFP isn't transmission to vendors. It's the act of articulating what your organization actually needs. Most failed system implementations begin with receiving vendor proposals without an RFP. Without a clear question, you end up buying what the vendor is best at selling."
"Three departments gave three different answers to the same question," I continued. "That is not a chatbot problem — it is a problem to be solved before the RFP is written. A chatbot is a tool for delivering information, but if the information being delivered isn't unified, it simply digitizes the silo."
"Let's use Between The Rows," Gemini proposed. "Enter the manuals and response guidelines from all three departments, and extract contradictions and gaps. Before deciding what the chatbot should learn, identify what's missing in the current information."
The manuals were entered. The analysis returned.
"Seventeen locations where phrasing conflicts between the three departments," Gemini read. "Twenty-three locations where missing information is likely generating inquiries. Once these are resolved, the scope of what the chatbot should answer becomes visible."
[The Four Pillars of an RFP]
"When writing the RFP, establish four pillars," Claude explained.
"Pillar one — define the problem to be solved in numbers. Reduce monthly inquiry response costs by sixty hours from the current hundred. By stating this number explicitly in the RFP, vendors understand the baseline for success. The foundation is quantifying the problem, not listing features."
"Pillar two — healthcare-specific constraints," Gemini continued. "Privacy law and medical information handling guidelines, integration requirements with existing hospital systems, device environments used by staff — without these, you'll receive a proposal for a generic chatbot. Healthcare implementation experience should be a selection criterion."
"Pillar three — operational structure for information updates," Claude continued. "A chatbot isn't a set-and-forget implementation. Every time a manual is updated, the chatbot's answers need updating too. Who across the three departments owns each update, and what is the update workflow — including this in the RFP means vendors design their after-sales support into the proposal."
"Pillar four — how to measure success," I concluded. "What is measured at three months to determine success? Reduction rate in inquiry volume, chatbot resolution rate, staff satisfaction — writing the measurement method into the RFP prevents misalignment with vendors."
Kawaguchi took focused notes. "I didn't realize there was this much to do before selecting a chatbot."
"The work of writing an RFP is the work of clarifying your own problems," I replied. "When problems are clarified, it becomes clear what kind of tool is needed. Inversely: decide on a tool before the RFP, and you buy a tool that doesn't fit the problem."
Chapter 3: Giving a Name to the Silent Cost
Kawaguchi organized his documents and said quietly:
"I once asked the staff in the three departments if they were feeling the strain of inquiry response. They all said 'we're fine.' But when I put it in numbers, a hundred hours a month was disappearing. For the first time, I feel like I've seen the silent cost."
"In healthcare settings," Claude added, "there's often a culture of not voicing workload. That's precisely why numbers generating questions matters. A hundred hours per month is also the most powerful justification for securing approval for improvement. The RFP is the bridge that translates that number into requirements for vendors."
"Next week, hold a workshop with the leaders of all three departments," Gemini proposed. "Review the conflict points extracted by Between The Rows together and agree on the correct information. That output becomes the information requirements for the RFP. Holding the workshop first means the operational update structure post-implementation also defines itself naturally."
Kawaguchi stood. "First, organize the information. Then convert it into the RFP and look for vendors. I'll move in that order."
Four months later, a report arrived from Kawaguchi.
The three-department information conflicts were resolved in the workshop and a unified manual was produced. An RFP was written and sent to five vendors. Selection criteria weighted healthcare track record and update support, resulting in one vendor selected.
Three months after chatbot launch, inquiry response volume dropped sixty-three percent. Staff surveys generated repeated comments: "I don't have to explain the same things over and over anymore."
Kawaguchi's report read: "Organizing the information before writing the RFP was the single greatest factor in success. With a unified manual complete, interdepartmental friction had already begun to decrease before the chatbot went live. It wasn't the tool that solved the problem — the process of preparing for the tool had already started solving it."
The day a tender with no question found its answer.
"Before asking a vendor for a proposal, there is the work of articulating what your organization is asking for. The RFP is the tool for that articulation. A tender without a clear question results in buying what the vendor is best at. When a hundred hours of silent cost is given a name, the question sharpens for the first time. A sharp question returns a sharp proposal. And only the side that holds the question is capable of evaluating the proposals."
Related Files
Tools Used
- Between The Rows — Automated extraction of contradictions and information gaps in manuals and guidelines
