A way of working where AI completes a task through a chain of steps it plans and checks itself, rather than one single reply.
In an agentic workflow the model is not asked to produce the final result in one shot. The task moves through stages instead: planning, executing, checking and correcting when needed. One model may play several roles, or several agents may each handle a specialty.
This approach produces better results on complex tasks like building a full app or analyzing a set of documents, because mistakes are caught and fixed along the way. The trade-off is more time and more tokens, so it is not the right fit for every simple task. Practical tip: when designing an agentic workflow in your app, add clear checkpoints between stages, such as an automated test or a human approval before any irreversible action like messaging customers or charging a card.