AI glossary Models & Architecture
What is an AI agent?
← All glossary termsSoftware that pursues a goal by deciding its own next steps, using tools and data, and acting on systems with limited human input.
Explained
An AI agent is a piece of software that receives a goal, works out the steps to reach it and carries them out. Today most agents use a large language model as the reasoning core: the model reads the situation, decides which tool to call (a search, an API, a database query, a calculation), looks at the result and decides what to do next. The loop repeats until the goal is met, a limit is reached or a person needs to step in. What makes it an agent rather than a script is that the path is not fixed in advance: the model chooses it at run time.
How an AI agent works
- Goal: a task stated in natural language or triggered by an event, such as a new email or a ticket.
- Reasoning: the model breaks the task into steps and decides which one comes next.
- Tools: functions the agent is allowed to call, such as search, read a document, update a record or send a message.
- Observation: the agent reads what each tool returns and adjusts the plan.
- Memory and context: notes on what has been done and what is still open, so long tasks do not lose track.
- Stop conditions: success, a step or cost limit, or a hand-off to a person.
Agent, chatbot or workflow?
A chatbot answers questions in a conversation; it does not usually act on other systems. A workflow runs a sequence of steps that someone designed in advance, even if some of those steps use a language model. An agent decides the sequence itself, within the tools and permissions it has been given.
In practice many useful systems mix the three: a fixed workflow for the predictable parts and an agent only where the path really depends on the case.
When it makes sense, and its limits
- It fits tasks with many possible paths that are hard to write as rules: triage, research, case handling, reconciliation.
- It needs clear tools, permissions and data; an agent cannot work around missing access.
- Each extra step adds latency, cost and room for error, so simple tasks are often better served by a workflow.
- Actions with real consequences (payments, customer messages, changes to records) usually need human approval, logging and evaluation before going live.
Frequently asked questions
What is an example of an AI agent?
A support agent that reads an incoming ticket, looks up the customer and order in the CRM, checks the returns policy, drafts a reply and opens a refund for a person to approve. It chooses which lookups to make depending on what the ticket says.
What is the difference between an AI agent and an LLM?
A large language model generates text from an input. An agent wraps a model in a loop with tools, memory and stop rules, so it can act on systems and work through several steps towards a goal. The model is the reasoning engine; the agent is the system around it.
Are AI agents the same as agentic AI?
They are closely related. An AI agent is a single system that plans and acts; agentic AI is the broader approach of building systems, often with several agents and workflows, that act with some autonomy rather than only answering prompts.
Is ChatGPT an AI agent?
On its own, a chat assistant mostly answers in conversation. When it is given tools, such as browsing, running code or acting in other applications, and can decide when to use them over several steps, it behaves as an agent.
Are AI agents safe to use in a company?
They can be, with the right controls: least-privilege access to tools and data, limits on steps and spend, human approval for sensitive actions, full logging and evaluation before and after launch. Most of the risk comes from what the agent is allowed to do, not from the model alone.
Decision guides that use this term
- Build vs buy AI agents: which agents should you own, and which should you rent?
- Agentic AI vs generative chatbots: does your use case need an agent or a good assistant?
- AI agents vs RPA: which one should automate your process, and when should you combine them?
- Microsoft Copilot vs custom AI agents: when is the suite assistant enough, and when do you need your own agents?
- Agent framework vs custom orchestration: should you adopt a framework or write your own agent loop?
- Human in the loop vs autonomous agents: how much should an AI agent decide on its own?
- Reasoning models vs standard LLMs: when is the extra thinking worth paying for?
- Single agent or multi-agent system: how should you architect your workflow?
How to put it to work
- How to automate accounts payable with AI
- How to automate email with AI
- How to automate processes with multiple AI agents
- How to automate the IT service desk with AI
- How to forecast cash flow with AI
- How to prepare sales meetings with AI
- How to red team your AI agents automatically
- How to use AI in procurement
Where it shows up at Thinkia
Further reading
Related terms
Sources
- Anthropic · Building effective agents
- Yao et al. · ReAct: Synergizing reasoning and acting in language models (arXiv)
- Wang et al. · A survey on large language model based autonomous agents (arXiv)
Updated: 2026-10-21