Guangdong has delivered a quantifiable scorecard for AI in government work. On October 21-22, the province's “Wanqing Government AI Bootcamp” convened in Guangzhou: over 100 officials from 22 pilot departments and 8 cities — plus delegations from Tibet's Nyingchi, which Guangdong supports — produced 54 working government AI scenarios in two days using “Wanqing WorkBuddy”, with average efficiency gains above 70% versus existing workflows. After re-verification, the scenarios will be replicated province-wide.
1. Real work, real materials: 54 scenarios in two days
The bootcamp was no concept demo. Participants were split into 10 teams, each with a dedicated scenario facilitator, with Tencent staff serving as instructors. Each team took one real business scenario through the full loop — scenario design, knowledge aggregation, skill building, expert assembly — then re-tested solutions with real case files.
Selected measured gains
① Initial review of IT project proposals: from 1 day to 15 minutes; municipal reviews from 2 days to 2 hours, with every comment traceable to its source clause;
② Pre-screening of service applications: from 15 minutes to 3 per case;
③ Archives Q&A across 100+ standards: from 10 minutes to under 1 per query;
④ Labor-arbitration case reports: from ~60 minutes to under 10;
⑤ Province-wide summary drafting: from 3-4 days to half a day; industry briefings across 65 sectors from ~64 hours to 8.
Review-and-approval scenarios were the most numerous; retrieval Q&A tackled “hard-to-find precedents, inconsistent interpretations”; drafting scenarios compressed document assembly. Guangdong's healthcare insurance bureau extended its maternity-allowance pre-screening into a Q&A-plus-review expert, cutting each work order from 30 minutes to 8.
2. Why replication works: a provincial hub with 15 million calls
Cross-department replication rests on a unified base. Wanqing is billed as China's first provincial-level government AI hub: since trial operation began in June it has logged over 15 million model calls, consumes 1 billion tokens daily, and runs 100 provincial knowledge bases with more than 370 million knowledge points. WorkBuddy runs on this shared foundation with province-unified entry, knowledge and security rules; it is deployed inside the government network, data stays in a locally controlled environment, the AI accesses files only within each user's existing permissions, and every operation is logged and traceable.
Selecting scenarios by frequency, time cost and reusability is what lets results travel — the same proposal-review scenario has already taken shape in the provincial bureau, Meizhou and Zhuhai, so replication does not start from zero.
3. From “innovate once, reuse everywhere”
Next, the provincial bureau will compile a panorama of Guangdong's government AI scenarios for distribution across departments. The ecosystem is expanding too: per education outlet Jiemodui, Tencent's Zhang Jun said on October 24 that over 100,000 teachers have claimed the free WorkBuddy edition for education.
The lesson generalizes beyond government: value comes from covering high-frequency, time-consuming, reusable tasks at scale. For enterprises and public institutions, the same formula — foundation model plus scenario methodology plus security boundaries — is what NineZenith's Tianxing LLM offers: private deployment, industry knowledge bases and full-process auditability that make AI upgrades measurable and replicable.
(Compiled from Guangzhou Daily, CNR and other public reports.)