Applying Large Language Models to Courier Operations: Intelligent Operations, Q&A, Prompting, and Agents

In 2022 OpenAI released ChatGPT, demonstrating impressive language understanding, content generation, and reasoning capabilities.

The courier terminal system, used by delivery personnel, benefits from large models to address operational issues, anomalies, and support needs, enabling tools for managers.

Analysis identified 143 courier actions, of which 69 have potential for LLM integration, such as data entry, outbound calls, SMS, tracking queries, and knowledge Q&A.

Intelligent operation uses voice input, ASR, LLM intent recognition, and API calls to automate tasks without separate model pipelines, reducing development effort.

Intelligent Q&A employs Retrieval‑Augmented Generation (RAG) to combine LLMs with external knowledge bases, improving answer accuracy and mitigating hallucinations.

Content extraction converts tables to semantic text; splitting respects paragraph boundaries; embeddings are generated with OpenAI models and stored in Vearch for similarity search.

Code examples illustrate document loading, table extraction, text splitting, and embedding generation:

from src.document_loader.document_loader import DocumentLoaderUtil
processor = DocumentLoaderUtil(file_path=path_ori, pic_save_dir=dir_save_picture)
texts = processor.load()
texts = json.dumps(texts, ensure_ascii=False, indent=4)
with open(os.path.join(dir_save_text, f"{os.path.basename(path_ori)}.txt"), "w") as f:
    f.write(texts)

from src.embedding.get_embedding import get_openai_embedding
model_key = "xxxx"
model_name = "text-embedding-ada-002-2"
texts_embedding = [
    get_openai_embedding(text=t.page_content, model_name=model_name, model_key=model_key)
    for t in texts_splitted
]

Smart prompts transform complex business rules into concise operational guidance, and LLM agents (GPT‑based) are explored for broader decision‑making tasks.