The Crisis of Entity Resolution in Heterogeneous AI Environments: Data, Financial, and Semantic Dimensions

Digital systems increasingly delegate the resolution of persons, organizations, products, places, and documents to language models, retrieval-augmented generation (RAG), and autonomous agents. Yet recognizing identical character strings is not equivalent to resolving identical entities. This study synthesizes research on entity linking, entity matching, temporally evolving knowledge bases, RAG, and multi-agent LLM systems with public enterprise data-quality surveys from 2020-2026. It demonstrates that identity decisions break down across four analytical zones: identity collapse under name ambiguity and popularity bias, semantic fragmentation across data silos, temporal discontinuity of changing entities, and the semantic vulnerability of retrieval context. These are not four isolated defects. In RAG architectures, they can cascade from indexing through retrieval to generation, whereas in agentic systems, an earlier misidentification can alter the target of the subsequent step, the next query, and even the purported verification. While economic evidence cannot be converted into a universal cost for a single entity error, it consistently documents the substantial operational scale of the problem: Gartner reports a historical average of at least USD 12.9 million per year attributable to poor data quality and predicts the abandonment of 60% of AI projects unsupported by AI-ready data by the end of 2026. In the Monte Carlo/Wakefield sample, two-thirds of respondents reported in 2024 a data incident costing at least USD 100,000 during the preceding six months; in the 2025 Fivetran/Redpoint sample, 42% of enterprises stated that more than half of their AI projects had been delayed, underperforming, or unsuccessful because of data readiness. The concluding theoretical section examines Daniel Beránek's Digital Identity Optimization (DIO) and Ontology of Digital Identity (ODI) as a novel framework for governing the integrity of digital identity in environments where the entity is reconstructed from fragmented, temporally variable, and mutually conflicting representations. Its empirical validation constitutes a subsequent phase of development, not a precondition for the existence of the problem itself or for the legitimacy of the framework. DIO is strongest precisely where the entity is not a single record, but a temporally, multilingually, and infrastructurally dispersed configuration of relations.