The short version
  • Drafting text, images, audio, code, and synthetic records are generative tasks.
  • Classification and forecasting are predictive tasks, even when AI performs them.
  • Many useful systems combine generation with retrieval and deterministic rules.

01

The simplest test

Ask whether the system produces a new artifact. Writing a response, generating an image, composing code, or creating a summary are generative tasks. Assigning a label or estimating a number is usually predictive.

02

Common generative task examples

Examples include drafting customer replies, transforming notes into a report, creating product imagery, producing test data, translating content, and generating code from a specification.

03

Tasks that are often confused with generation

Search, anomaly detection, lead scoring, recommendation ranking, and demand forecasting do not inherently generate content. They may still be part of a broader generative AI workflow.

Common questions

Questions worth asking

Is summarization a generative AI task?

Yes. The model creates a new, shorter representation of source material, even though it is constrained by existing content.