AI agent
An AI agent is a system that is given a goal and a set of tools and then decides for itself which tool to call, in what order, and when it has finished, instead of following a route that a person wrote down in advance; the same case can take a different path on two different runs.
Submit a project Choose agent or workflowReviewed by Huy Nguyen (David), CEO · Updated 15 Oct 2026 · 3 min read
This term is one of the entries in the OutsourcingVN glossary of words buyers meet while scoping an AI feature's level of autonomy.
Why does an AI agent matter?
The word "agent" gets attached to almost any AI feature in 2026, which makes it hard to know what you are actually buying. OutsourcingVN is operated by Netbase JSC, and the distinction that actually matters for a buyer is not how advanced a system sounds but who chooses the next step — software following a fixed route, or the model itself deciding what to do next — because that choice decides how predictable, testable and auditable the system will be.
What is an example of an AI agent in practice?
A research task asks a system to find out why a competitor's pricing page changed last month. An AI agent is given a web-browsing tool and a note-taking tool and decides on its own which pages to open, when to search again, and when it has gathered enough to write a summary. Two runs over the same question might open a different number of pages in a different order, because nobody wrote the exact path down in advance; a person still reads the summary before acting on it.
What are the common misreadings of an AI agent?
An AI agent is not simply a chatbot with a longer conversation — a chatbot that follows a fixed menu of intents is a deterministic workflow wearing a conversational interface, not an agent, because the software still decides the branch. It is also not inherently more capable than a fixed pipeline; for a process that can be written down in advance, a pipeline is usually more predictable and cheaper to test. And calling a system an agent does not excuse it from evaluation: an agent's output still needs a held-out test set, even though its path cannot be enumerated the way a pipeline's can.
Plan the next step for your project
Common questions
No. A chatbot that follows a fixed menu of intents and hands off to a fixed branch is a deterministic workflow; an agent decides its own next step and tool, which a fixed-branch chatbot does not do.
Where its actions are reversible or a person checks the output before it counts, review can happen after the fact; where an action writes to a customer record or a production system, a gate belongs in the path before the action runs.
Which decisions does an AI agent affect?
Choosing between an AI agent and a deterministic pipeline with one bounded model step is the subject of the AI agent vs workflow automation guide, which compares predictability, testing, failure modes and change cost; the emerging technology adoption guide covers the same bounded-experiment method for an agent alongside any other new AI capability. When you are unsure whether a process needs an agent's flexibility or a fixed route would serve it better, describe the process, and Netbase will respond through Netbase's own outsourcing-services platform; submit a project when you are ready.