AI agent, AI assistant or chatbot: what is the difference and which one to choose
Three words that vendors use as synonyms. What stands behind each, where they do not replace one another and which one to start with
When one contractor writes "AI agent" and a competitor offers a "chatbot", a business owner often sees no difference. There is one, and the launch time, the risks and the budget depend on it. Below is what each word means in practice and how to choose for your task. If you already know what you need, you can go straight to AKORDO AI solutions for sales and customer service.
Chatbot: works from a script
A chatbot leads a person along pre-written branches: buttons, ready answers, a booking form. It does not make things up, so its behaviour is easy to predict and check. It has one weak spot: when the customer writes in their own words, the bot gets lost or answers off the point.
It fits when there are few questions, they repeat and every branch of the conversation fits on one sheet of paper.
AI assistant: understands words and answers from a knowledge base
An AI assistant takes a free-form question and builds an answer from company material: regulations, price lists, service descriptions. It talks, but it does nothing in your systems: it does not create a deal, change a status or send an invoice. If the answer is not in the base, the right behaviour is to say so and call a person.
It fits first-line support and questions about a product, delivery or service rules. The catalogue has this as AI customer support 24/7.
AI agent: understands and acts in systems
An AI agent does not only answer, it takes steps: checks data in the CRM, creates a task for a manager, updates a customer card, passes the request on by the rules. For that it gets access to systems and clear limits on what it does itself and what it only proposes to a person.
So an agent needs more preparation: described processes, access rights, an action log and a test on sample data. An example from the catalogue: the AI sales manager 24/7 answers new enquiries, qualifies the lead and passes it to the CRM.
Comparison
| Feature | Chatbot | AI assistant | AI agent |
|---|---|---|---|
| How it answers | Ready branches and buttons | From a knowledge base, in free words | From a knowledge base and system data |
| Acts in systems | No, or minimally | No | Yes, within agreed limits |
| What to prepare | A conversation script | An answer base and handover rules | A process, access, rules, an action log |
| Main risk | The customer is stuck in a branch | An inaccurate answer | A wrong action in a system |
| Used for | Simple repeated questions | Support and consultation | Routine between a channel and the CRM |
Market names are blurry, so with a contractor do not ask "what type is this". Ask three things: where the answer comes from, what the system does by itself and what happens when it does not know.
How to choose
- Describe the task in one sentence. "Customers ask about delivery" and "leads hang between the site and the CRM" are different tasks.
- If the task is about answers, start with an assistant or even a scripted bot. It is cheaper and quicker to check.
- If the task is routine between systems, look at an agent, but start with one scenario, not the whole business.
- Before launch, decide who owns quality: who reads the dialogues, who updates the base, who gets the hard question.
If you are not sure where to start, an AI audit and quick start shows which processes to hand to AI and which to leave with people. In the AKORDO catalogue it is from $600 to $1 000 and from 2 to 4 weeks, the exact sum depends on scope.
What to read next
For the choice between a scripted bot and an assistant, what a launch includes and questions for a contractor, see A chatbot for business: what to choose. For a first practical scenario for an agent, see AI agent: the first scenario. You can discuss your task at an AKORDO consultation.