Agentic Workflow
An agentic workflow is a process where an AI model drives multiple steps toward a goal — planning, calling tools, and reacting to results in a loop — rather than producing a single response. It sits between a one-shot prompt and a fully autonomous agent: structured enough to be reliable, dynamic enough to handle real tasks.
Also known as: agentic workflows, agent workflow
A single prompt does one thing. An agentic workflow chains the model’s decisions across steps toward a goal: it plans an approach, calls tools to act, reads the results, and decides what to do next — looping until done. It’s the operational shape of an AI agent, and the unit most production “agents” actually are.
The design tension is autonomy versus reliability. A rigid, fully scripted pipeline is predictable but can’t adapt; a model free to do anything adapts but is hard to trust. Most teams that ship land in between — letting the model decide within bounded steps, with guardrails and evaluation at each stage. That middle ground is exactly where a lot of the practical agent-building on the show happens: enough structure to be dependable, enough flexibility to be useful.
Go deeper
From the conversation
-
Got Agents? Agentic Workflows & Architecture | Weaviate, Unstructured & CrewAI - The Making of Gemini 2.0: DeepMind's Approach to AI Development and Deployment | Logan Kilpatrick
- Practical Lessons for GenAI Evals | Chip Huyen & Vivienne Zhang
- Hallucinations Are a Data Architecture Problem | Sudhir Hasbe, Neo4j
- Mastering Multi-Agent Systems | MongoDB’s Mikiko Chandrasekhar
-
The First Fully Autonomous AI Attack Is 18 Months Away | Kristin Lovejoy -
The AI Framework Era Is Over: Why Context Is the Moat | Jerry Liu - Beyond Transformers: How Liquid AI Is Rethinking LLM Architecture | Maxime Labonne