"Chatbot" and "AI agent" are used for very different products, and software companies do not always use the words the same way. The difference matters when you choose one, because it decides how much control you have over what the software says and does, and how much work it takes to keep it right. This guide explains the main types of chatbots, what makes something an AI agent, and how to decide which one your business needs.
If the job you have in mind is answering customer questions, our guide to what a support chatbot should and should not handle covers that job from start to finish.
The short answer
A chatbot holds a conversation. An AI agent holds a conversation and also acts: it uses an AI model to decide what to do next, and it can carry out tasks in other software, such as looking up an order or changing a booking, without a person scripting each step.
Every AI agent is a kind of chatbot, but most chatbots are not AI agents. A chatbot can follow a script the business wrote, use AI to answer questions, or do both.
The main types of chatbots
The parts of a short lead conversation
- Message. The greeting says who you are and what the chat can do.
- Buttons. The common answers are buttons, so the visitor taps instead of typing.
- Collect Input. The bot asks for one detail and saves it to the lead.
- Add Tags. The visitor never sees this step. It labels the lead by its answers, so you can sort your Leads inbox.
- Message. The closing line says what happens next.
Rule-based chatbots
A rule-based chatbot, also called a scripted, flow or decision tree chatbot, follows a conversation that someone designed in advance. It shows a message, offers buttons or asks a question, and moves to the next step based on the answer. It says only what was written for it.
Its strength is that it is predictable. Every visitor gets the same questions in the same order, so a lead's details are collected the same way every time and nothing is made up. Its limit is that it cannot answer a question nobody wrote an answer for.
Keyword chatbots
A keyword chatbot looks for particular words in what the visitor types, such as "price" or "hours", and replies with the answer attached to that word. It handles typed questions better than buttons alone, but it is easily confused by wording it was not set up for, such as "what do you charge".
AI chatbots
An AI chatbot uses a language model, the kind of AI that writes text, to reply in its own words. It is given information to answer from, such as a page of business details, help articles or documents, and it can answer the same question asked in many different ways.
The drawback is less control over what it says. An AI chatbot chooses its own wording, so it needs accurate information, a clear instruction about what to do when it does not know, and someone who reads its conversations.
Hybrid chatbots
A hybrid chatbot uses a script for the conversations that must go the same way every time, such as booking an appointment or collecting a lead's details, and AI for the open questions in between. For many small businesses this is the practical middle ground: the important paths stay predictable, and a visitor with an unusual question still gets an answer.
AI agents
An AI agent does more than answer. It can use tools, which means taking actions in other systems: checking an order's status, changing an appointment, issuing a refund or updating an account. The AI decides which action to take and when, within limits the business sets, and it can work through several steps to finish a task.
That makes agents useful for businesses with many systems and a large number of routine requests. It also adds risk. An agent that can change a booking can change the wrong one, so agents need careful limits, thorough testing and working connections to the software they act on.
Live chat
Live chat is not a chatbot at all. A person on your team answers the visitor in real time, and many tools combine it with a chatbot that answers first. Our comparison of chatbots and live chat covers when a person in the chat is worth the cost.
AI agent vs chatbot, side by side
The three types people most often compare differ in how they decide what to say, what they can change, and what you have to look after.
| Rule-based chatbot | AI chatbot | AI agent | |
|---|---|---|---|
| How it decides what to say | Follows the script you wrote | Writes its own reply from the information it has | Writes its own reply and decides which action to take |
| What it can change | Only what you built into the script, such as saving a lead or booking a time | Nothing; it answers questions | Records in the other systems it is connected to |
| What you look after | The script | The information and the instructions | The information, the instructions, the connections and the limits |
| Main risk | A question nobody scripted | A confident wrong answer | A wrong action |
| Best suited to | Lead capture, booking and set questions | Varied questions about one business | Large numbers of routine requests across several systems |
Which one does your business need?
Start from the job rather than the technology. Four questions usually settle it.
- What do people contact you about? If most messages are the same ten questions plus requests for appointments or quotes, a rule-based or hybrid chatbot covers them. Automating the questions you answer every day shows how to start.
- Do the questions vary widely? If customers ask about hundreds of products, policies or situations, AI answering is worth adding.
- Does the chatbot need to change anything in your other software? If customers mainly want something done, such as a refund or a change to an order, that is work for an agent, and it needs the software and the staff to support it.
- Who will look after it? A script needs an occasional hour. An AI chatbot needs someone to read its conversations. An agent needs someone who understands every system it touches.
For many small businesses, the answer is a scripted chatbot first, with AI added once the conversations show questions the script does not cover.
Where Alma fits
Alma is a hybrid in the sense above: a flow chatbot with optional AI answers. You build the conversation in a drag-and-drop flow builder, usually starting from a template, and the chatbot follows it on your website with messages, buttons, questions and tags. Alma also has its own booking calendar, with a booking page the chatbot can link to.
AI answering is an optional add-on for $25 a month, with up to 500 AI answers a month. It answers typed questions your flows do not cover, using only the business information you provide as text, up to 50,000 characters. It does not read your website or uploaded files, and the Alma team sets it up and loads your text into the bot. The free FAQ generator turns your business details into text of that kind.
Alma is not an AI agent. Its AI answers questions but does not take actions. The actions in an Alma chatbot, such as adding a tag or sending details to another system with an API Call step, are steps you place in a flow, so they happen only where you put them. Alma also has no live chat: when a visitor needs a person, the chatbot can ask for an email address or phone number, and Alma emails the lead to the address you set for notifications.
If you are weighing Alma against an AI agent platform or a help desk, our comparison pages set out what Chatbase, Intercom, Fin, Tidio, Botpress and Drift do, using each company's own website.
Common questions
What is the difference between an AI agent and a chatbot?
A chatbot holds a conversation. An AI agent also takes actions in other software, such as changing a booking or looking up an order, and uses AI to decide which action to take. Every AI agent is a chatbot, but most chatbots are not agents.
What are the main types of chatbots?
Rule-based chatbots that follow a script, keyword chatbots that react to particular words, AI chatbots that write their own replies, hybrid chatbots that combine a script with AI, and AI agents that can also take actions. Live chat, where a person replies, is often offered alongside them.
Is a rule-based chatbot out of date?
No. For booking, lead capture and set questions, a script is often the better tool, because every visitor gets the same questions and nothing is made up. AI is most useful for the open questions a script cannot predict.
Does a small business need an AI agent?
Only if customers mostly want something changed in your other software and you can connect the agent to it. If they mainly want answers, prices and appointments, a flow chatbot with optional AI answering covers the job.
To see what Alma's flows and AI answering do in more detail, read the features page. When you are ready, start the 14-day free trial and build the flow first.
