A basic assumption is that effective artificial intelligence must be closely embedded in, and articulated with, the daily workflow of the organization. This chapter presents a model of how that embedding and articulation might be achieved in a nuclear power plant, with a specific focus on equipment maintenance. The model describes the sociotechnical organization within a nuclear power plant, includes multiple nested control layers, and provides context for describing the communication and coordination challenges associated with safe and efficient operations. A significant component of the model, essential for achieving workflow embedding and articulation, is a framework for coding the results from a corpus of data that includes basic maintenance reports. In the nuclear industry, these are referred to as condition reports. The coding process described in the framework provides a mechanism for extracting meaning (semiotics) from the corpus of data. This semiotic framework addresses both the correspondence problem, ensuring that the data accurately reflect the physical and technical processes and associated safety envelopes involved in nuclear power, and the coherence problem, ensuring that the data are comprehensible to the relevant human decision-makers. Together, the sociotechnical model and semiotic framework provide an essential set of mechanisms for making sense of the large corpus of maintenance reports generated by the nuclear industry. These mechanisms provide a pathway for integrating artificial intelligence in nuclear power plants via existing retrieval augmented generation (RAG) architectures. Such architectures require the kind of domain-specific authoritative databases curated by domain experts that the sociotechnical and semiotic frameworks can provide. This chapter outlines the research basis for the sociotechnical and semiotic framework. This research describes how the practical analysis of maintenance failure issues, using a sociotechnically based multilevel organizational control structure, provides a basis for a meaning-based semiotic coding structure. This structure provides the mechanisms for integrating the RAG architecture/artificial intelligence with an existing sociotechnical and organizational system in the nuclear sector. The result has the potential of enhanced problem-solving and decision-making. The ultimate goal is to ensure safe and efficient plant performance in line with principles for human-centered artificial intelligence.