Transforming Operations with Generative AI and Workforce Optimization

1. Revamping Accounting Operations with Generative AI

Implementation Process

• Development of an Invoice Data Extraction System Using RPA
   In just three weeks, an automated system was built to extract data from invoices, significantly reducing manual data entry errors.
• Development of an AI-Enhanced Expense Report Verification Feature
   Over a two-month period, a verification function was developed that incorporates generative AI to automatically check expense reports, enhancing both accuracy and the speed of the approval process.
• Training a Budget Optimization Algorithm
   Utilizing three years’ worth of financial data, a budget optimization algorithm was trained to refine budget planning and improve strategic decision-making.

Numerical Outcomes

Task Before After Reduction Rate Monthly Closing 72 hours 24 hours 67% Expense Processing 60 hours/month 25 hours/month 58% Audit Preparation 220,000 hours/year 187,000 hours/year 15%

This initiative significantly enhanced the efficiency of the accounting department, enabling a reallocation of resources towards more strategic tasks.

2. Implementing the “Skill Puzzle” AI Tool for Workforce Optimization in Manufacturing

Optimization Algorithm and Evaluation Metrics

Workforce optimization in manufacturing goes beyond simple task assignment; it simultaneously considers several metrics including:

• Skill Match Score
   Evaluates the percentage of essential qualifications held by employees for each process.
• Workload Distribution Index
   Reduces the proportion of heavy work time per employee, easing overall workload.
• Experienced Personnel Allocation Balance
   Increases the placement ratio of veteran workers in key processes to ensure quality and facilitate skills transfer.

Case Study: Automobile Parts Manufacturer

• Production Efficiency
   Daily production increased from 320 units to 378 units, an 18% improvement.
• Defect Rate
   Reduced dramatically from 2.1% to 0.9%.
• Workforce Scheduling Time
   The time required to create daily staffing plans dropped from 8 hours to just 45 minutes.

Additionally, a new trend emerged in February 2025 with Fujifilm’s announcement of an AI model that quantifies tacit knowledge of skilled workers. This model learns machine operation patterns, allowing the expertise of veteran technicians to be effectively transferred to younger employees. The system has demonstrated significant benefits in both skills retention and optimal workforce allocation.

3. Deploying an AI Analysis System in Call Centers

The KDDI Case Study

In call center operations, the key challenge is to efficiently analyze a vast volume of customer interaction notes to drive service improvements. KDDI achieved remarkable results:

• Data Processing Volume
   The system automatically processes 150,000 interaction records per month.
• Classification Accuracy
   Achieved a high accuracy rate of 92.3%, comparable to human judgment.
• Report Generation Time
   The time to generate analytical reports was reduced from 40 hours per month to just 2.5 hours, enabling quicker decision-making.

System Architecture

The system is structured as follows:

1. Input of Interaction Records
2. Data Storage in Amazon S3
3. Classification Processing via Bedrock
4. Creation of Category-Specific Databases
5. Automated Generation of Trend Analysis Reports

This streamlined process ensures that customer feedback is captured accurately and rapidly analyzed, leading to tangible service quality improvements.

4. Economic Impact Analysis of Workforce Optimization AI

Workforce optimization not only boosts operational efficiency but also delivers clear economic benefits. For a factory with a workforce of 300 employees, the projected benefits include:

• Annual Labor Cost Reduction
   Approximately 48 million yen saved annually.
• Revenue Increase Through Enhanced Productivity
   An estimated increase in revenue of 120 million yen.
• Savings in Training Costs
   Annual savings of 6.5 million yen.

Based on these figures, with an estimated initial investment of 35 million yen, the payback period is approximately 14 months. Key considerations during implementation include:

1. Preprocessing of internal data (digitizing three years’ worth of operational records)
2. Clear definition of evaluation metrics (KPI establishment for ROI calculation)
3. A phased rollout (pilot phase followed by full-scale deployment)