Study goals
Compare a chiller maintenance plan developed using the traditional method with another generated by AI (ChatGPT and DeepSeek), evaluating accuracy, reliability, flexibility, cost, implementation time, decision-making and regulatory compliance.
Relevance / originality
Explores, for the first time, the practical use of generative AI in asset maintenance planning, applying Bloom’s Taxonomy and detailed comparative analysis, expanding efficiency possibilities and integration with predictive technologies in Facilities Management.
Methodology / approach
Comparative case study between three methods: traditional, ChatGPT and DeepSeek. Used prompts structured by Bloom’s Taxonomy and evaluated six performance dimensions based on experts, including technical parameters, cost-effectiveness and regulatory compliance.
Main results
ChatGPT showed greater flexibility, organization and technical detail. DeepSeek excelled in operational risk-based analysis. Both outperformed the traditional method in some criteria but require human validation to ensure regulatory compliance and safe practical application.
Theoretical / methodological contributions
Integrates Bloom’s Taxonomy into AI use in maintenance, providing a replicable methodology for comparing generative tools. Offers a conceptual basis for future studies on integrating predictive maintenance and AI in asset management.
Social / management contributions
Highlights AI’s potential to reduce costs, optimize time and support more assertive decision-making in building maintenance, strengthening asset management modernization and improving operational reliability in corporate environments.