Agentic AI
Agentic artificial intelligence (AI) describes systems that use AI to pursue goals across multiple steps, choosing actions and adapting to their results. Usage varies: some authors also include workflows whose steps are largely predefined.
An order-support system might look up a shipment, notice missing tracking data and decide to query a second service before answering. The action loop connects observation, a choice of next step and the result of that step. The ReAct (reasoning and acting) paper demonstrates interleaving model reasoning with actions that obtain information from an environment.
The label does not specify how much freedom a system has. Anthropic uses “agentic systems” broadly, then distinguishes predefined workflows from agents that dynamically direct their own process. Describe the actual control flow: which decisions belong to the model, which belong to code, and which require a person.
This matters because a system can do more than produce a wrong sentence: it can choose a wrong action repeatedly. Set permissions, stopping conditions and checks on completed work. Artificial general intelligence concerns breadth of capability; agentic AI concerns how a system acts toward a goal. One does not establish the other.
Sources
- Yao et al.: ReAct — Demonstrates interleaving reasoning and actions with environmental feedback.
- Anthropic: Building effective agents — States its broad use of agentic systems and distinguishes workflow control from agent control.
Go deeper
- What's the difference between agentic and non-agentic AI? AI, decoded · Agentic vs. Non-Agentic AI
- What makes an AI agent different from an LLM? AI, decoded · 4 Things That Turn a Model Into an Agent
- Hugging Face: Agents guided tour docs
Follow an agent through model-selected tools, observations and a stopping condition.
From the conversation
-
The First Fully Autonomous AI Attack is Coming | Kris Lovejoy, Kyndryl -
The Agent Bubble Debate | Spot AI's Kelly Vaughn - Breaking the Language Barrier: Smartling's AI Translation Pipeline | Olga Beregovaya
- How Block Deployed AI Agents to 12,000 Employees in 8 Weeks w/ MCP | Angie Jones
- Inside IBM's watsonx: Building Enterprise AI That Ships | Dr. Maryam Ashoori
-
Tuning GPU Performance with AI Agents | AMD’s Anush Elangovan on ROCm 10 -
Every AI Agent Has an Evaluation Gap | Alex Ratner, Snorkel AI - How AI Velocity is Rewriting the Rules for Engineering Leaders | ChatPRD's Claire Vo