In the midst of a global digital transformation, corporate financial management is undergoing profound changes. With the rapid advancement of artificial intelligence (AI), the emergence of large-scale models like DeepSeek has brought new possibilities to financial digitization, helping enterprises transition into an era of intelligent financial management.
AI in Action: A Case Study from AVIC Qingyun
Since February 2025, the Finance Department of AVIC Qingyun, in collaboration with the Planning & Development Department and the Production & Operations Department, has been exploring ways to deeply integrate DeepSeek and other large AI models into financial digitization processes.
After multiple trials with DeepSeek’s model and Alibaba’s Qwen2.5-VL32B (a large-scale vision-language model, hereafter referred to as “Qwen Model”), the team ultimately adopted Qwen due to its visual recognition capabilities—an area where DeepSeek’s current model was not yet optimized. Leveraging massive GPU computing power for large-matrix calculations and deep learning-based reasoning, the AI successfully:
-
Extracted key information from thousands of mixed-content invoice images.
-
Automated invoice recognition & data extraction, processing over 4,000 backlogged invoices.
-
Built an invoice image & information database, significantly improving efficiency in pricing support documentation and providing robust support for product pricing reviews and R&D expense audits.
A Cross-Departmental AI Collaboration
This initiative marked a multi-departmental effort to enhance enterprise management through AI:
-
Finance Department: Proposed the management vision, objectives, and coordination.
-
Planning & Development Department: Provided hardware resources and technical guidance.
-
Production & Operations Department: Handled Qwen model adaptation and debugging to implement the solution.
Beyond invoice processing, the team explored broader AI applications, including:
-
Building a local private knowledge base.
-
Developing AI agent tools based on the MCP protocol for customer service automation.
The human-computer interaction (HCI) design outperformed traditional OCR software in:
-
Feature refinement & mining.
-
Scene layout comprehension.
-
Element & positional relationship analysis.
-
Key information classification & extraction.
The Road Ahead: Expanding AI in Financial Digitization
As the project progresses, AVIC Qingyun plans to further explore AI applications in financial digitization. This groundbreaking experiment also sets a precedent for other departments to adopt AI models, accelerating the company’s journey toward intelligent management innovation.
By integrating large AI models’ generalization capabilities and human-like problem-solving skills, enterprises can unlock unprecedented efficiency and accuracy in financial operations—ushering in a new chapter of smart finance.