AI agents · Types
What types of AI agents can a company build?
The short answer
A company builds four types of AI agent: conversational agents, workflow agents, autonomous agents and multi-agent systems. They differ in two things: what starts the agent, and who decides its next step. Between them they take ten forms.
| Type | Forms |
|---|---|
| Conversational agents | Answering agent, acting agent |
| Workflow agents | Scheduled agent, event-triggered agent, API agent |
| Autonomous agents | Autonomous agent |
| Multi-agent systems | Agent pipeline, manager agent, agent team, guardian agent |
Why not the usual five?
Ask what types of AI agent exist and most sources give the same five: simple reflex, model-based, goal-based, utility-based and learning agents. IBM's guide lists them, and they come from the standard university textbook on artificial intelligence.
That list is accurate. It describes how an agent reasons. It does not say whether a person or a system starts the agent, whether it may change anything, how many agents are involved, or who approves. Those are the first decisions a team makes, and they set the cost and the risk.
The ten forms in one line each
- Answering agent. Answers questions from knowledge you have approved and shows the source. It changes nothing in any system.
- Acting agent. Answers and also carries out the request through the tools it has been given.
- Scheduled agent. Runs at set times, or over a batch of items, and produces the same kind of output each run.
- Event-triggered agent. Starts when something happens in a system: an email arrives, a file lands, a record changes.
- API agent. Another application calls it, it does one job, and it returns a result in a fixed format.
- Autonomous agent. Is given a goal, not steps. It plans, acts, checks the result and adjusts.
- Agent pipeline. The builder fixes the order in advance. Each agent does its part and passes its output to the next.
- Manager agent. One agent decides, while the work is under way, which specialist handles each part, and assembles the result.
- Agent team. Specialists hand work directly to one another. No single agent is in charge.
- Guardian agent. Its job is to watch, check or stop other agents.
How do you choose?
Six questions settle it.
- Does a person ask each time? If yes, it is conversational. If the answer ends the matter, an answering agent. If something must then be done, an acting agent.
- If nobody asks, what starts the work? A clock: scheduled. Something happening: event-triggered. Another system: API.
- Can you write the steps down in advance? If you can, stay with a workflow agent or a pipeline. If you cannot, an autonomous agent, with limits.
- Is the work too wide for one agent? The signs are instructions that pull against each other, or more tools than one agent chooses between reliably.
- Who should decide the order? You, in advance: a pipeline. One agent: a manager. The specialists themselves: a team.
- What happens if it is wrong? If the answer involves a customer, money or a regulated record, add an approval step and a guardian.
Then take the simplest form the answers allow. OpenAI, Anthropic and Microsoft give the same advice: begin with the simplest design and add complexity only when the simple one falls short.
Which jobs use which form?
| The job | The forms usually used |
|---|---|
| Staff questions on policy, HR and IT | Answering agent first, acting agent for requests |
| Customer service | Acting agent. An agent team where cases move between specialisms |
| Regulatory and market monitoring | Scheduled agent for the daily read. A pipeline to assess each change |
| Intake of claims, invoices and applications | Event-triggered agent. An API agent for a single decision |
| Research and investigation | Autonomous agent |
| Operations and incidents | An event-triggered start, with a manager agent to coordinate |
| Anything sent to a customer or a regulator | A guardian agent before it leaves, and a person who approves |
Where should a company start?
Start with an answering agent on your own documents. No system is changed, so it can run on the first day, and it shows whether your knowledge is in a state an agent can use.
Then take one process and one form: an acting agent for a single type of request, or an event-triggered agent for a single type of document. Run it beside the manual process and measure it against that.
Add a second agent when the first can no longer hold the work, and not before.
Questions people ask
Is a chatbot an AI agent?
By the strict definitions, no. OpenAI's guide says applications that use a model without letting it control the work, simple chatbots among them, are not agents. An assistant that only answers from documents is the boundary case. We count it as the first form because buyers ask for it by the name "agent" and most organisations start there.
What is the difference between a workflow agent and an autonomous agent?
In a workflow agent the builder sets the job in advance and the agent handles each item. An autonomous agent is given a goal and plans its own steps.
When do you need more than one agent?
When one agent's instructions pull against each other, when it has more tools than it chooses between reliably, or when parts of the work belong to different teams or security boundaries.
What is a guardian agent?
An agent whose job is to watch, check or stop other agents. Gartner names three roles: reviewers that check output, monitors that track actions, and protectors that adjust or block an action as it happens.
Does every business problem need an agent?
No. Microsoft's guidance says so plainly. If the steps are fixed, ordinary automation is cheaper and more predictable.