.svg)
With mobile FIN recognition and comprehensive LLM evaluation, AMAI supports AVAG Holding SE speeding up processes and building your AI strategy on a secure technical foundation.
challenge
AVAG Holding faced two key challenges of digital transformation: On the one hand, the optimization of an everyday but error-prone process in car dealerships — the manual recording of chassis numbers (VIN). On the other hand, the strategic question of how modern language models (LLMs) can be profitably and securely integrated into corporate IT in order to tap into future automation potential. We were looking for a pragmatic technology partner who could deliver tangible prototypes and well-founded technical recommendations quickly and without detours.
solution
In close, partnership-based collaboration, we developed two tailor-made AI applications that are directly tailored to AVAG's needs:
- FIN recognition prototype: We designed and implemented a streamlined web application that allows car dealership employees to simply photograph the VIN of a vehicle with their smartphone. An intelligent pipeline of OCR technology (paddleOCR) and specific post-processing algorithms extracts the 17-digit number from the image, automatically validates it for the correct format and outputs it for further use.
- Strategic LLM Evaluation: In order to make the potential of large language models tangible for AVAG, we carried out a detailed technical analysis. Using load tests, we compared self-hosting approaches with cloud solutions (Azure OpenAI) and demonstrated the performance of the technology using a specific use case of automated email generation for expiring leasing contracts. The result is a clear, data-based recommendation for AVAG's future AI strategy.
core features
- Mobile capture of chassis numbers directly on the vehicle using the photo function.
- Automatic text recognition (OCR), optimized for detection on vehicle windows.
- Intelligent post-processing to correct typical OCR errors (e.g. letter “O” instead of number “0") and validate the FIN structure.
- Profound load tests to compare various LLM hosting options for robust infrastructure
- Automated generation of high-quality, personalized texts such as e-mail messages for customers.
AVAG benefits
- Significant time savings and reduction of manual entries for car dealership employees.
- Minimize entry errors and thus an increase in data quality in downstream systems.
- Clear, technical basis for decision-making for the strategic use of LLMs in companies.
- Demonstration of specific automation potential, e.g. in customer communication, to relieve staff.
- Quick and agile implementation from idea to working prototype, which received very positive feedback from end users.
outlook
Following the extremely successful test phase of the FIN recognition prototype, AVAG is planning a company-wide rollout in car dealerships. The findings gained from the LLM evaluation form the basis for the gradual introduction of AI-supported text generation, starting with the automated creation of follow-up offers for leasing customers. Both use cases have the potential to be integrated even more deeply into existing core processes in the future and to further increase efficiency throughout the company.
customer feedback
“Working with AMAI was a real win for us: fast, uncomplicated implementation, absolutely professional appearance and impressively deep expertise. I particularly appreciate the excellent communication — on equal footing, transparent and always solution-oriented. This is what partnership-based and efficient cooperation looks like to me.”
Bernd Ottens
Chief Information Officer at AVAG Holding SE
