On 11 September the Jingjiang Tax Bureau in Jiangsu ran a training session on large models and AI agents, focused on teaching staff to build their own agent-based assistants. The demonstrations stayed close to daily work: looking up policy rules quickly, analysing risk indicators, polishing written reports and sorting scattered data. The session also drew hard lines for internal AI use - no classified or sensitive data may be uploaded, key information must be de-identified first, AI output must always be reviewed by a human, and tax data must stay inside the intranet. As the bureau put it, efficiency gains have limits and tax data is never a small matter. Officials said they will keep hunting for suitable scenarios and turn the training into front-line results. Attendees included bureau leaders, heads of departments and sub-bureaus, and staff aged 40 or under.
News Detail
Jingjiang Tax Bureau Runs Training on Large Models and AI Agents for Tax Work
On September 11 the Jingjiang Tax Bureau in Jiangsu held training for its staff on large language models and AI agents, combining theory with hands-on practice and demonstrating policy lookups, risk-point analysis, report drafting and data collation. The session also set hard rules for internal AI use, including no uploading of classified or sensitive data, de-identification before use, mandatory human review of AI output and keeping tax data inside the internal network. The bureau said it will keep exploring agent use cases and turning the training into front-line results; leadership, department heads and staff aged 40 or under took part.
2026-09-24