Glossary

Generative AI

Generative AI is a class of artificial-intelligence systems designed primarily to produce new content — such as text, images, audio, code, or other structured outputs — conditioned on prompts, context, or other inputs. Contemporary generative systems are often built on foundation models (including large language models), but generative AI is not identical to AI as a whole, to predictive/discriminative AI, or to agentic systems that plan and act with tools.


Why it matters

Enterprise programs frequently label every model initiative “generative AI,” which collapses distinct design and risk choices. Generative systems change the work pattern from classifying or forecasting to drafting and synthesizing artifacts at scale — which raises questions of quality, IP, hallucination, evaluation, and workflow redesign that differ from classical predictive models.

Useful distinctions: predictive AI estimates labels or scores; foundation models are broadly adaptable pretrained bases; LLMs are a major text-centric foundation-model family; generative AI names the content-producing application class; agentic systems add goal-directed action loops over tools and environments. A chat assistant may be generative without being a full agent.

Hypothetical illustration: A product organization deploys a generative drafting assistant for customer emails using a foundation-model API with retrieval. Separately, a fraud team continues to run a predictive risk-scoring model. Treating both as the same “AI program” obscures different evaluation metrics, failure modes, and governance controls — generative quality and hallucination risk versus precision/recall on fraud labels.

Mechanisms

  • Conditional generation — models sample or decode outputs from a learned distribution given prompts and context.
  • Adaptation stack — prompting, fine-tuning, RAG, and tools specialize generative behavior for enterprise tasks.
  • Quality and risk surface — factuality, IP leakage, toxicity, and cost attach to generated artifacts and must be evaluated as such.

Boundaries

  • Generative AI ≠ all of artificial intelligence; predictive and optimization systems remain distinct.
  • Generative AI ≠ foundation model or LLM as exact synonyms (those name model classes/layers that often power generative applications).
  • Generative AI ≠ AI agent / agentic AI; generation alone does not imply multi-step tool-using agency.
  • Access to a generative model is not by itself an AI strategy or a durable advantage.

Sources

  1. Bommasani, Rishi, et al. — On the Opportunities and Risks of Foundation Models (Stanford CRFM) — foundation models as bases commonly used for generative applications.
  2. Goodfellow, Ian, et al. (2014). Generative Adversarial Nets — https://arxiv.org/abs/1406.2661 — early influential generative-modeling line (GANs); contemporary enterprise generative practice is broader than GANs alone.
  3. Brown, Tom B., et al. (2020). Language Models are Few-Shot Learners — https://arxiv.org/abs/2005.14165 — large language models as a dominant generative text paradigm.



Further research


Continue exploring

A short path from this concept into related research and practical resources.

  1. Related conceptFoundation model
  2. Deeper researchWhy Ai Features Dont Compound
  3. Apply in DiscoverThe Enterprise AI Playbook Library
Generative AI — Definition for Strategists | The Strategy Stack