A company has terabytes of structured data and wants an AI application that converts employee natural-language requests into SQL queries. Employees have limited technical skills. Which solution best meets this need?

Generative pre-trained transformers (GPT) are best suited for this task because they excel at understanding natural language and generating coherent, contextually relevant text, including code like SQL. Their ability to process and generate human-like text makes them ideal for converting natural language requests into structured queries, especially given the employees' limited technical skills. A residual neural network is primarily used for image recognition and deep learning tasks, not natural language to code generation. A support vector machine is a supervised learning model used for classification and regression, not generative text or code. WaveNet is a deep neural network for generating raw audio, not text or SQL queries.