What Is an AI Agent?
Also called agentic AI, agentic
Short Answer
An AI agent is a system where a language model works toward a goal in a loop: it decides the next step, calls tools such as search, code or business systems, looks at the results and continues until the task is done. Unlike a chatbot, it takes actions, not only answers.
- Taught in lesson 113
- In 10 real job ads
Why It Matters
Agents are how AI moves from answering to doing: researching, filling forms, handling tickets or changing code. Companies are investing heavily in them, so the ability to design agents that are useful, safe and testable is one of the most sought-after AI skills.
How It Works
A useful distinction comes from Anthropic's 2024 guide "Building effective agents": a workflow follows steps fixed in code, while an agent lets the model choose its own steps. Many reliable products are workflows with agent-like parts.
Agents bring new risks: they can loop, take wrong actions or be misled by text they read. Production agents use limited tool permissions, approval steps for risky actions, stopping conditions, tracing and evaluation of whole task runs.
An agent combines a model, a set of tools with clear descriptions and permissions, a loop that runs until the goal is reached or a limit stops it, and often memory. Because agents take actions, design choices matter: keep steps predictable where possible, limit what tools can do, log every step, ask a person before irreversible actions and test on realistic tasks. Many successful agents are narrow and structured rather than fully open.
AI Agent in Practice
Everyday Example
Asked to book a meeting, an agent checks three calendars, finds a free slot, drafts the invitation and asks you to approve before sending it.
Industry Example: Pharma
A pharmacovigilance agent gathers a case's reports from several systems, drafts a case summary and routes it to a safety physician, who approves or corrects it.
Where You Meet It
- Coding assistants that edit and test code.
- Research assistants that search and summarise.
- Back-office automation that updates systems.
- AI engineer job ads mentioning agents or agentic workflows.
Key Ideas Around It
- Tool
- A function the agent can ask to run.
- Loop
- Decide, act, observe, repeat.
- Memory
- Information the agent keeps across steps.
- Human approval
- A person confirming risky actions.
A Common Mix-Up
More autonomy is not always better. A predictable workflow with a few tools is often more reliable and cheaper than an agent that plans everything itself.
Check YourselfWhat separates an agent from a chatbot?
An agent takes steps and uses tools to act toward a goal; a chatbot only replies to messages.
Seen in Real Job Ads
Lines quoted word for word from live job postings.
Sanofi
Digital R&D Staff Software Engineer, Hyderabad
Understanding of agent memory models (short-term, long-term, episodic), tool use, and planning strategies (ReAct, Plan-and-Execute, etc.)
propio
AI/LLM Safety Engineer, Overland Park, KS
Perform threat modeling for agentic scenarios: tool-call boundaries, sandbox isolation, and least-privilege access, with particular attention to preventing agents from exfiltrating data or executing irreversible actions through chained tool calls.
propio
AI/LLM Safety Engineer, Overland Park, KS
Design and maintain a safety evaluation framework: adversarial prompt sets, scenario-based test suites, and regression suites, so that every model and agent update is validated before it ships.
Learn It on ai2hired
S10 · #113Workflows vs Agents
Coming in Season 10, Agents and Tool Use. Every ai2hired lesson is free on YouTube.
Questions People Ask
What is the difference between an AI agent and a chatbot?
A chatbot answers in conversation. An agent pursues a goal across several steps and acts through tools.
Are AI agents safe to use at work?
They can be, with limited permissions, human approval for risky actions, logging and testing. Without those, they are risky.
Related Terms
Learn AI From Zero to Hired
Free lessons on YouTube, each one built from what real AI job ads ask for.