Glossary · AI
Agentic workflow
Also called: AI workflow, LLM workflow, AI-powered workflow
An agentic workflow is a workflow with a fixed overall structure in which some steps use a language model to read, classify, decide between predefined branches or draft content.
The term sits between two poles. On one side, a classic workflow: every step and branch is written by a person. On the other, an AI agent: the model chooses its own steps. An agentic workflow keeps the route map human-designed and lets AI do the parts rules are bad at, such as reading messy input, classifying intent or writing a draft.
Common patterns
- Extract, then act. AI turns an invoice or email into fields; ordinary steps do the rest.
- Classify and route. AI labels a ticket "billing, urgent"; a router sends it to the right queue.
- Draft and approve. AI writes the reply; a person approves before it's sent.
- Evaluate and retry. One AI step drafts, another checks the draft against criteria, and the workflow loops a bounded number of times.
Why most business automation should start here
Every step is visible and testable. Cost is predictable: you know roughly how many model calls a run makes. Failures are contained: a bad classification goes down a wrong branch you can review, instead of an agent wandering off. And the AI parts can be swapped (another model, another prompt) without redesigning the whole thing.
Start with an agentic workflow. Promote a step to a full agent only when you find yourself writing dozens of branches for paths you can't predict.
In Fluxo Flow
Most Fluxo Flow workflows are agentic workflows: an AI Generate step for a single prompt (with JSON output), a Router on its result, a Validator against a JSON schema, and Human Review where the stakes are real. The Agent node is there for the steps that need tool-calling. Templates like AI lead qualification and support ticket triage are agentic workflows you can open and Dry Run.
Read next
Related terms
- AI agentAn AI agent is a system in which a language model decides its own next step, calls tools (APIs, searches, other systems), reads the results and repeats until it judges the goal reached.
- WorkflowA workflow is a repeatable sequence of steps, started by a trigger, that turns an input (a form, an invoice, a new deal) into an outcome, with defined rules for each decision along the way.
- Structured outputStructured output is a model response in a predefined, machine-readable format, typically a JSON object matching a schema, so the next steps of a workflow can use the fields directly instead of parsing prose.
- Human in the loop (HITL)Human in the loop (HITL) is a design in which an automated process pauses at chosen points for a person to review, correct, approve or reject its output before it continues.
- ActionAn action is a workflow step that does something outside the workflow, such as creating a CRM record, sending an email, posting to Slack or calling an API.