The real reason enterprise AI is stuck

Enterprise AI is currently limited by its reliance on human-centric metaphors rather than formal computational models, according to an analysis of the industry's development. While metaphors like "memory," "reflection," and "planning" aid in understanding and communication, they do not facilitate industrial-scale deployment. This reliance on analogy, exemplified by Business Insider's description of Anthropic's "dreaming" technique for AI agents, contrasts with the formalization required for industrialization. The article posits that major software revolutions, such as relational databases and the World Wide Web, followed a pattern of initial capability, followed by formalization, and then the emergence of a platform. For instance, relational databases became industrial after Edgar F. Codd introduced a formal relational model, which later enabled SQL and a broader ecosystem. Similarly, the World Wide Web's transformative impact stemmed from the formal identification of resources, as defined by the W3C's Architecture of the World Wide Web, rather than just improved browser aesthetics. This distinction between metaphor and formal model is presented as the core reason why enterprise AI struggles to move beyond impressive demonstrations to reliable, widespread deployments.