Document Processing
Kingsoft Office wanted to launch a domestic AI office product quickly, but needed a Chinese-first large-model stack, strong content-safety controls and an operating model for continuously tuning prompts and training data across office scenarios.
Reported impact
Kingsoft launched a domestic AI office assistant that covers multiple content formats, improves user productivity and created an iterative quality-and-safety operating model around office AI.
Kingsoft Office connected WPS AI to Baidu AI Cloud's Qianfan model platform to create a broad document- and presentation-assistance layer across WPS Office and WPS 365. The technical pattern is a multi-loop operating model rather than a simple API connection. First, an AI-technology middle layer handles large-model API integration, parameters, error codes and quality monitoring. On top of that, prompt engineering defines roles, instruction scopes and chain-of-thought patterns that adapt the model to office tasks. A second loop uses labeled bad cases and SFT data from text, slides and spreadsheets to keep improving output quality. Together those inner and outer loops let Kingsoft co-build and tune office scenarios such as drafting, rewriting, summarizing, polishing, translation and PPT generation, while Qianfan's content-safety controls form a safety boundary around production use. The result is a model-driven office copilot that serves multiple file types and creation stages rather than a single feature.
User document, slide and spreadsheet requests flow through WPS AI into the Qianfan model stack, where prompt engineering and labeled training data optimize generation quality before results are returned to office applications.
Kingsoft connected WPS AI to Baidu's Qianfan platform and quickly launched the first AI office application in China's collaborative-office market (original: 第一时间通过百度智能云千帆大模型平台接入了文心一言，快速推出了 “WPS AI”，这是中国协同办公赛道首个智能办公应用).
The delivery model includes API integration, prompt engineering and labeled-data loops for continuous optimization (original: 第一步是 AI 技术中台接入……第二步是 Prompt 提示词工程……第三步是精标数据驱动).
WPS AI covers drafting, rewriting, summarizing, polishing, translation and PPT generation (original: 包括起草、改写、总结、润色、翻译、续写等……WPS AI 可协助从 PPT 大纲生成、PPT 制作、模板选择、内容修改的全过程).
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
Baidu AI Cloud Sample May 15, 2025 Wuzhi Technology (悟智科技) Wuzhi needed an automated way to turn large volumes of paper-based and highly varied documents into accurate, searchable, and structured digital assets for government and enterprise customers without long manual… Document retrieval time decreased 40%. Graph Database · Hybrid Search and Multi-route Retrieval · Intelligent OCR 
Google Cloud Sample Modified · Oct 9, 2026 Anthropic During large-scale training runs, traffic is highly variable—dropping to near zero during model checkpointing before bursting to millions of events per minute. Anthropic needed a serverless messaging queue that scaled… Handles peak throughput of over 1M events per second BigQuery · Bigtable · Cloud Pub/Sub 
NVIDIA Sample Sep 22, 2026 Sluicebox Sluicebox needed to turn fragmented bills of materials, supplier records, PDFs, CAD files, and emissions inventories into timely product carbon assessments, but manual life-cycle-analysis work was slow, expensive, and… 51-80% lower inference costs Lucy · NVIDIA Nemotron 3 Ultra · carbon graph 