Generative AI

Generative AI refers to models that produce new content, text, images, audio, code, rather than assigning a label or a number to existing input, which is what most earlier machine learning systems did. A large language model is the dominant form of generative AI for text, but the category also includes image, audio, and video generation models built on related deep learning architectures.

The shift from predictive to generative models changes what "output" means: instead of a category or score, the system returns open-ended content that must be judged on qualities like coherence, relevance, and factual accuracy rather than simple correctness against a label. This makes evaluation harder and makes hallucination a distinct concern that classification models did not have.

Generative AI matters to analytics because it turns data work that used to require manual writing or design, drafting summaries, generating synthetic data, building an AI copilot for analysts, into something a model can do directly from a prompt. The common pitfall is applying generative tools to problems that need guaranteed accuracy or reproducibility, such as financial reporting, without validation; generative output should be checked with the same rigor as any other automated process, typically through careful prompt engineering and downstream review.