ROI Case File No.379 | 'TechSavvy Inc.'s Company-Wide Wall of AI Adoption'

Chapter 1: The Trap of Partial Success—The Invisible Wall of Company-Wide Deployment

The day after resolving TechNexus Inc.'s OKR case, a consultation arrived regarding company-wide AI adoption deployment. Volume 30, "The Pursuit of Reproducibility," Case 379 tells the story of separating problem discovery from solution exploration.

"Detective, we have success stories. The Marketing department utilized ChatGPT and reduced SNS post creation time by 60%. The Sales department also uses Claude Projects to automate initial draft generation for proposals. However, we can't deploy company-wide. And integration with core systems (SAP and OBIC's HR/payroll management systems) is difficult because we share them with our parent company."

Misaki Takahashi, DX Promotion Office Director of TechSavvy Inc. from Shinagawa, visited 221B Baker Street with an expression full of anxiety. In her hands were AI utilization success case reports from Marketing and Sales departments (20 pages each), and in stark contrast, a company-wide deployment plan titled "AI Implementation Roadmap 2026-2027."

"We are an IT solution company. 380 employees. Annual revenue of 9.5 billion yen. However, AI utilization is limited to some departments. For next fiscal year implementation, we aim to enable department-wide adoption by receiving various proposals."

TechSavvy Inc.'s Current State: - Founded: 2008 (IT solutions) - Employees: 380 - Annual revenue: 9.5B yen - AI utilization status: Marketing (15 people), Sales (25 people) only - Issues: Can't deploy company-wide, difficult core system integration, unclear specific implementation locations

Deep anxiety permeated Takahashi's voice.

"The particular challenge is visitor reception. Currently, administrative staff respond every time, interrupting their work. Average 15 visitors per day. Average 8 minutes response time per visitor. 5 administrative staff spend 150 hours monthly on this alone. We want to automate this, but don't know which AI tools to use."

Visitor Reception Reality:

Current Flow: 1. Visitor presses entrance intercom 2. Administrative staff responds ("Yes, who is calling?") 3. Visitor explains visit purpose (average 2 minutes) 4. Staff makes internal call to relevant department (average 3 minutes) 5. Department contact person comes to reception (average 5 minutes) 6. Staff hands visitor to contact person (average 3 minutes) Total time required: 13 minutes/visitor (8 minutes staff response + 5 minutes wait)

Monthly Performance: - Visitors: 300/month (20 business days × 15/day) - Staff response time: 300 × 8 minutes = 2,400 minutes (40 hours) - Administrative staff: 5 people (rotate responses) - Monthly response time per staff: Average 8 hours

Visitor Breakdown: - Business meetings: 180 (60%) - Sales visits: 60 (20%) - Job interviews: 30 (10%) - Other (deliveries, inspections): 30 (10%)

Takahashi sighed deeply.

"There's another problem. Which operations should receive AI isn't clear. Reception automation is one example, but there should be other applicable operations. However, we don't have time or resources to survey all 380 employees' work. And because we have core systems shared with our parent company, we're considering introducing independent AI tools."

Chapter 2: The Trap of Solution-First Thinking—The Problem's Essence Is Invisible

"Takahashi-san, do you believe implementing AI tools will automate reception?"

Takahashi answered immediately.

"Yes, I expect that. I thought an AI chatbot could ask visitors questions and automatically transfer them to the relevant department."

Current Understanding (Solution-First Approach): - Expectation: Solve everything at once with AI chatbot - Problem: The problem's essence is invisible

I explained the importance of separating problem discovery from solution exploration.

"The problem is solution-first thinking: 'implementing AI tools will solve it.' Double Diamond—a Design Thinking process model. Separate problem space exploration (1st Diamond) from solution space exploration (2nd Diamond). First discover 'what problem should truly be solved,' then explore 'what's the optimal solution,' achieving reproducible value creation."

⬜️ ChatGPT | Catalyst of Concepts

"Don't put solutions first. Separate problem discovery from solution exploration with the Double Diamond"

🟧 Claude | Story Alchemist

"Problems are always 'true challenges hidden behind visible symptoms.' Discovery first, solution second"

🟦 Gemini | Compass of Reason

"Follow the Double Diamond's 4 stages. Discover, Define, Develop, Deliver"

The three members began their analysis. Gemini displayed the "Double Diamond" on the whiteboard.

Double Diamond Structure:

[1st Diamond: Problem Space] Diverge → Converge Discover → Define[2nd Diamond: Solution Space] Diverge → Converge Develop → Deliver

Each Stage: 1. Discover: Broadly explore problems (diverge) 2. Define: Identify essential problems (converge) 3. Develop: Broadly explore solutions (diverge) 4. Deliver: Implement optimal solution (converge)

"Takahashi-san, let's start by discovering the essential problem with reception."

Chapter 3: 1st Diamond—Problem Space Exploration

Phase 1: Discover - Broadly Explore Problems (2 weeks)

Survey Methods:

Survey 1: Administrative Staff Interviews (5 people) - Question: "What troubles you about reception?" - Time: 1 hour each

Survey 2: Visitor Questionnaire (100 people) - Question: "What would you like improved about reception?" - Method: Online questionnaire via QR code

Survey 3: Observation Study (Shadowing) - Period: 5 days (business days) - Method: Observe and record administrative staff reception responses

Discovered Problems (Diverge):

Administrative Staff Voices: 1. "Work is interrupted each time visitors come" (all 5 people) 2. "Often don't know who the visitor is for" (4 people) 3. "When contact person is absent during internal call, don't know how to respond" (all 5 people) 4. "Delivery personnel responses are included, but that's a separate issue" (3 people) 5. "Can't grasp visitor schedules in advance" (all 5 people)

Visitor Voices (100 questionnaire respondents): 1. "Wait time from reception until contact person arrives is long" (78 people) 2. "Even after explaining who I'm visiting, sometimes get passed around" (42 people) 3. "Even though I made an appointment, I need to explain again" (65 people) 4. "Reception was unmanned at times, which was troublesome" (15 people)

Observation Study Findings: 1. Administrative staff leave their seats 3-5 times daily for reception 2. When contact person is absent during internal call, average 4 transfers occur 3. Visitor schedule table (Excel file) isn't updated in 60% of cases 4. For 20% of sales visits (walk-ins), relevant department isn't immediately known

Phase 2: Define - Identify Essential Problem (1 week)

Converge Discovered Problems:

Problem Grouping:

Group 1: Work Interruption Problem - Administrative staff work is interrupted - 3-5 times daily, total about 40 minutes

Group 2: Information Sharing Problem - Visitor schedule table not updated (60%) - Don't know who visitor is for - No determined response for absent contact persons

Group 3: Wait Time Problem - Long visitor wait times (average 5 minutes) - Pass-arounds occur (42%)

Essential Problem Identification (Converge):

Problem Statement: "Because visitor schedule information isn't shared in advance and confirmation is needed each time at reception, administrative staff work is interrupted and visitor wait times are also long"

Problem Metrics: - Staff interruption count: 15 times/day - Staff response time: 40 hours/month - Visitor wait time: Average 5 minutes - Visitor schedule table update rate: 40%

Chapter 4: 2nd Diamond—Solution Space Exploration

Phase 3: Develop - Broadly Explore Solutions (2 weeks)

Solution Divergence:

Idea 1: AI Chatbot Reception - Visitors answer questions on tablet - AI identifies relevant department and auto-notifies

Idea 2: Pre-Registration System - Visitors pre-register visit information via QR code - Just scan QR code at reception on the day

Idea 3: Facial Recognition System - Auto-identify regular visitors with facial recognition - Auto-notify relevant department

Idea 4: Smart Lock + AI Voice Response - Visitors voice input via intercom - AI understands visit purpose through voice recognition and auto-notifies

Idea 5: Integrated Calendar System - Link with internal calendars - Automatically display visitor schedules on reception screen

Evaluation of Each Idea:

Evaluation Axes: - Axis A: Implementation cost (1-5 points, lower is higher score) - Axis B: Effect (work interruption reduction rate) (1-5 points) - Axis C: Visitor convenience (1-5 points)

Evaluation Results:

Selection: Adopt Idea 2 (Pre-registration + QR Code)

Phase 4: Deliver - Implement Optimal Solution (2 months)

Month 1: Prototype Development

System Design:

Visitor-Side Flow: 1. Pre-register visit schedule via online form 2. Input visit date, time, department/contact person, visit purpose 3. QR code is generated (sent via email) 4. On the day, scan QR code at reception 5. Automatically notify relevant department (Slack integration)

Administrative Staff-Side Flow: 1. Check visitor schedule list on reception tablet 2. When QR code is scanned, automatically send Slack notification to relevant department 3. Manual internal call only in emergencies

Contact Person-Side Flow: 1. Receive visitor notification in Slack 2. Tap "Heading to reception" button 3. Automatically display "Contact person is on their way" to visitor

Technology Stack: - Frontend: React + TypeScript - Backend: Node.js + Express - Database: PostgreSQL - Notifications: Slack API - QR code generation: qrcode.js

Development Cost: - System development: 2.8M yen - Tablet purchase (for reception): 80K yen - Total initial cost: 2.88M yen

Month 2: Pilot Operation and Effect Measurement

Pilot Period: 1 month

KPI Measurement:

KPI 1: Pre-Registration Rate - Visitors: 300/month - Pre-registered: 240 (80%) - Day-of registration: 60 (20%)

KPI 2: Staff Response Time - Before: 40 hours/month (300 × 8 minutes) - After: 8 hours/month (60 × 8 minutes, no response needed for 240 pre-registered) - Reduction rate: 80% - Time saved: 32 hours/month

KPI 3: Visitor Wait Time - Before: Average 5 minutes - After: Average 1.5 minutes (for pre-registered) - Reduction rate: 70%

KPI 4: Slack Notification Response Time - Until contact person taps "Heading to reception" button: Average 30 seconds - Traditional internal call: Average 3 minutes - Improvement rate: 83%

KPI 5: Visitor Satisfaction - Survey conducted (100 people) - "Wait time became shorter": 92 (92%) - "Was smooth": 88 (88%)

Month 3 Onward: Company-Wide Deployment and Continuous Improvement

Improvement 1: Add AI Chatbot (for day-of registration) - For visitors without pre-registration - AI chatbot asks questions on tablet - Estimates relevant department from visit purpose and sends Slack notification

Improvement 2: Integrated Calendar Integration - Link with Google Calendar - Automatically display visitor schedules on reception screen - Auto-verify with contact person's schedule

Annual Effects:

Personnel Cost Reduction (Administrative Staff): - Time saved: 32 hours/month × 12 months = 384 hours/year - Personnel cost reduction: 384 hours × 3,200 yen = 1.23M yen/year

Productivity Improvement (Administrative Staff): - Interruption count reduction: 15 times/day → 3 times (80% reduction) - Productivity improvement from increased focused work time: Estimated 20% - Effect: Approximately 2M yen/year

Visitor Time Value: - Wait time reduction: 5 minutes → 1.5 minutes (3.5 minutes saved) - Monthly 300 visitors × 3.5 minutes = 1,050 minutes (17.5 hours)/month - Annual reduction: 210 hours - Visitor hourly rate conversion: 5,000 yen - Time value: 1.05M yen/year

Total Annual Effects: - 1.23M + 2M + 1.05M = 4.28M yen/year

Investment: - Initial development: 2.88M yen - Annual maintenance: 600K yen

ROI: - (4.28M - 0.6M) / 2.88M × 100 = 128% - Payback period: 2.88M ÷ 3.68M = 0.78 years (9.4 months)

Chapter 5: Detective's Diagnosis—Separating Problem Discovery from Solution Exploration

That evening, I contemplated the essence of the Double Diamond.

TechSavvy Inc. had solution-first thinking: "implementing an AI chatbot will solve it." However, through the 1st Diamond (problem space exploration), the essential problem became visible.

Problem Definition: "Because visitor schedule information isn't shared in advance and confirmation is needed each time at reception, administrative staff work is interrupted and visitor wait times are also long"

From this problem definition, in the 2nd Diamond (solution space exploration), we diverged 5 ideas and converged with evaluation axes. As a result, we discovered the optimal solution: "Pre-registration + QR Code."

What's important is that we separated problem discovery from solution exploration. Solution-first approaches miss essential problems. First broadly discover problems (Discover) and define the essence (Define). Then broadly develop solutions (Develop) and deliver the optimal solution (Deliver).

Staff response time reduced 80% (40 hours → 8 hours), visitor wait time reduced 70% (5 minutes → 1.5 minutes), annual effects of 4.28M yen, ROI of 128%, payback in 9.4 months achieved.

"Don't put solutions first. Separate problem discovery from solution exploration with the Double Diamond. Discover, Define, Develop, Deliver. By following these 4 stages, reproducible value creation emerges."

The next case will also depict the moment of discovering the problem's essence.

"Double Diamond—Discover and define problems in the 1st Diamond, develop and deliver solutions in the 2nd Diamond. By separating problem discovery from solution exploration, true value emerges"—From the Detective's Notes

Related Files