This edition of the Data Pro News provides an extensive enterprise reality check on the deployment of GPT-5, two months after its release, noting a growing gap between its initial promise and production reality. It explains that while the model offers impressive capabilities, key challenges remain, including spiralling infrastructure costs due to massive memory and compute requirements for its large context window. The article highlights that cost management for token usage at scale is proving prohibitive for many organisations, often leading to six-figure monthly expenditures. Furthermore, the reliance on Retrieval-Augmented Generation (RAG) persists, as using the full context window is neither cost-effective nor performant, meaning successful adoption requires robust engineering practices and hybrid architectures rather than simply integrating the new model. The text concludes that future enterprise AI success will depend less on access to advanced models and more on the engineering capability to deploy and operate them efficiently.
