Types of AI Agents: The Classic 5 and Today's Business Types
Short answer
There are five classic types of AI agents: simple reflex, model-based reflex, goal-based, utility-based and learning agents. In business today, agents are usually grouped by job (support, sales, coding, research), by interface (chat, voice, browser, background) and by autonomy, from assistants that suggest to agents that run whole jobs.
What are the 5 types of AI agents?
The five classic types of AI agents come from the textbook Artificial Intelligence: A Modern Approach by Stuart Russell and Peter Norvig. They are ranked from simplest to most capable.
- Simple reflex agents act on the current input with fixed rules. Example: a thermostat, or an email filter that moves messages with a certain word.
- Model-based reflex agents keep an internal picture of the world, so they can act on things they cannot see right now. Example: a robot vacuum that remembers which rooms are clean.
- Goal-based agents choose actions that move toward a goal. Example: a navigation app planning a route.
- Utility-based agents pick the best option among several that reach the goal, by scoring trade-offs like time and cost. Example: a ride-share pricing system.
- Learning agents improve from feedback over time. Example: a recommendation engine that adjusts to what you click.
Where do LLM agents like ChatGPT fit in these types?
Modern LLM-based agents are mostly goal-based agents with some learning ability. You give them a goal, and they plan steps to reach it using tools.
They also keep a model of the situation through memory and context, and many score options before acting. In practice, a product like Claude Code or Intercom Fin mixes several classic types. That is why business buyers use different groupings, below.
What types of AI agents do businesses use?
Businesses usually group AI agents by the job they do. This is the most useful way to shop, because it maps to a budget line and a team.
- Front office: AI receptionists, voice agents and customer service agents.
- Growth: sales agents, AI SDRs, marketing and SEO agents.
- Build: coding agents and agent builders.
- Operations: bookkeeping, meeting assistants, project management and research.
What are the types of AI agents by interface?
AI agents also differ by where they work: in a chat window, on the phone, in a browser or in the background with no interface at all.
- Chat agents work in a website widget, Slack or WhatsApp. Example: Intercom Fin.
- Voice agents talk on the phone. Example: Smith.ai, Synthflow, agents built on Vapi or Retell AI.
- Browser and computer-use agents click and type in websites or apps like a person. Example: ChatGPT agent, Perplexity Comet, Claude for Chrome.
- Background agents run on triggers with no chat. Example: a Zapier or n8n agent that processes every new invoice.
What are the levels of AI agent autonomy?
AI agents range from assistants that suggest to agents that run a whole job unsupervised. We rate every agent on this site on three levels.
- Level 1, assists you: drafts, suggests and summarizes; you take the action. Example: Microsoft 365 Copilot drafting a reply.
- Level 2, does tasks you hand it: completes a task you assign, then reports back. Example: ChatGPT agent booking a restaurant.
- Level 3, runs whole jobs: handles a full workflow on its own, escalating only exceptions. Example: an AI receptionist answering every call.
What is a multi-agent system?
A multi-agent system is a group of AI agents with separate roles working on one task. One might plan, another research, another write and another review.
Developers build these with frameworks such as CrewAI, LangGraph, Microsoft's Agent Framework and the OpenAI Agents SDK. Most small businesses never need one; a single well-scoped agent covers most jobs. Learn more in how to build an AI agent.
Frequently asked questions
What are the 4 main types of AI agents?
+
Some sources list four: simple reflex, model-based reflex, goal-based and utility-based. The fifth, learning agents, is often added because any of the four can also learn from feedback.
What type of agent is ChatGPT?
+
In agent mode, ChatGPT is a goal-based agent that uses tools like a browser and code interpreter. In normal chat mode, it behaves more like a conversational assistant than an agent.
What is a hierarchical agent?
+
A hierarchical agent setup has a manager agent that splits a task and hands parts to worker agents, then combines their results. It is a common pattern in multi-agent systems.
Which type of AI agent is best for a business?
+
Pick by job, not by textbook type. Start with the task that costs you the most time, such as calls, support or outreach, and choose an agent built for it.
What is a vertical AI agent?
+
A vertical AI agent is built for one industry, such as Harvey for law firms or Abridge for medical notes. It comes with industry knowledge and integrations a general agent lacks.