A pipeline for AI-generated monthly cost-analysis reports: content framework, data sources, and a four-step generation process
AI-generated monthly cost analysis reports = content standards + data pipeline + automatic calculation + attribution + LLM drafting: the content skeleton follows the three elements of RISN-TG 012-2011 Clause 8.3.2 (cost-indicator completion / profit-loss cause analysis / proposed measures), expanded per the 13 dynamic-management items of DBJ/T 15-153-2019 Clause 7.7.3; the four-step pipeline is automatic data extraction and cleaning -> three-way comparison + five-dimension analysis + earned-value calculation -> quantity/price/change attribution -> LLM template-based drafting, relying on budget, actual-consumption, price, contract, and variation / site instruction / certified record ledger data foundations.
Applicable Codes
- RISN-TG 012-2011 建设工程项目经理执业导则
- DBJ/T 15-153-2019 广东省建设项目全过程造价管理规范
- DB42/T 823-2021 建设工程造价咨询质量控制规范
- GB/T 50326-2017 建设工程项目管理规范
- GB/T 50358-2017 建设项目工程总承包管理规范
- GB/T 51235-2017 建筑信息模型施工应用标准
- T/ZBD 100.1-2022 建设工程造价数据建设通用标准
Topic Positioning
This card covers how to use AI to automatically generate project monthly cost-analysis reports: stringing report content standards, data pipeline, automatic calculation, variance attribution, and LLM drafting into an actionable four-step pipeline, as a reference for digital cost-management implementation.
Core Conclusions
The report content has a normative skeleton: RISN-TG 012-2011 Clause 8.3.2 requires the report to contain three elements — cost-indicator completion, profit-loss cause analysis, and proposed measures; DBJ/T 15-153-2019 Clause 7.7.3 refines the dynamic cost-management report into 13 items, the core being "the difference between the currently forecast project cost and the approved estimate/investment-control target" and its deviation-cause analysis. The AI pipeline has four steps: (1) automatic data extraction and cleaning (from the construction-drawing budget, actual labor/material/machinery consumption database, material price database, contract ledger, and variation / site instruction / certified record and claim ledger); (2) automatic calculation: three-way comparison (budget/plan/actual) + five-dimension analysis (time nodes, task-breakdown units, organizational units, individual indicators, comprehensive project, GB/T 50326-2017, Clause 11.5.2) + earned-value four indicators (BCWS/BCWP/ACWP, CV=BCWP-ACWP, SV=BCWP-BCWS, GB/T 50358-2017, Clause 2.0.8); (3) variance attribution: quantity variance = (actual qty - budget qty) x budget price, price variance = (actual price - budget price) x actual qty, variation and certified site record increments attributed to scope changes; (4) LLM generates the final report from templates.
Applicable Boundaries
Applies to monthly (or milestone-based) cost analysis of construction projects that have budget/planned cost baselines and established consumption/price/contract ledger data foundations; without the data foundation, AI reports are "cooking without rice". The five dimensions and earned value apply only to priced calibers with baselines; auto-generated attribution conclusions still require cost-staff review. Not applicable to scattered questions like single-point price queries.
Further Reading
See RISN-TG 012-2011 "Guidelines for practicing construction project managers" Clause 8.3.2; DBJ/T 15-153-2019 "Specification for whole-process cost management of construction projects in Guangdong" Clauses 7.7.3, 3.7.4; DB42/T 823-2021 "Specification for quality control of construction cost consultancy" Clause 9.6.2; GB/T 50326-2017 Clauses 11.5.1, 11.5.2; GB/T 50358-2017 Clause 2.0.8; GB/T 51235-2017 Clause 9.3.5; T/ZBD 100.1-2022 Clauses 4.2.6, 5.2.8 (cost database construction).
Keywords
- 月度成本分析报告
- AI自动生成
- 成本分析
- estimate-budget-final account comparison
- earned value
- 量价差归因
- 造价动态管理报告