Back to Case Studies
Used truck distributor
Client: Truck Ohkoku Co., Ltd.
Truck Condition Inspection Agent
A digital intake inspection agent that replaces tacit visual judgment with image AI
This case study covers how Optimium Inc. built a truck condition-inspection AI agent for Truck Ohkoku Co., Ltd., the company behind Truck Ohkoku (トラック王国), a Japanese used-truck marketplace. The agent standardizes intake appraisal that previously depended on veteran appraisers' eyes and instinct.
Project overview
| Item | Detail |
|---|---|
| Client | Truck Ohkoku Co., Ltd. — operator of the Truck Ohkoku marketplace |
| Timeline | ~5 months (2 months of data prep and model validation + 3-month build) |
| Team | 2 Optimium engineers (ML engineer / engineer) |
| Phases | Requirements → Data prep & model validation → Production build → Operational tuning |
| Key tech | Per-damage-type image recognition models, LLM-generated findings, appraiser-override feedback loop |
Problem
Condition appraisal for incoming used trucks was driven by veteran appraisers' eyes and instinct. With appraisers becoming harder to hire and reliance on individual judgment causing inconsistent standards, the time-to-resale and the accuracy of refurbishment cost estimates had plateaued.
Approach
- Trained per-damage-type discriminative models (exterior scratches, bed dents, frame corrosion, paint chips, body-mount wear) on the client's historical intake photos and appraisal records
- Combined image analysis with an LLM so the agent presents not just a verdict but where on the truck — and what observation — led to it
- Built an operations loop where appraiser overrides feed back into training data, so accuracy compounds on the lot
Outcome
- Standardized appraisal criteria — even new appraisers deliver consistent quality
- Higher appraisal throughput and more accurate refurbishment estimates
- Auditable reasoning paired with photos improves traceability for post-sale claims