A chatbot answers a message. An agent runs the journey.
The two words get used interchangeably, and the difference is not marketing. It decides whether a conversation ends in an order or in somebody repeating themselves to a human.
Here is what actually separates them, in terms you can check against any vendor.
Six differences that matter when you are buying
Ask about each of these. The answers tend to be quite different from the brochure.
- Memory across days
- A chatbot usually treats each message as a fresh start. An agent holds the thread, so a customer who comes back tomorrow is not asked everything again.
- Resuming a half-finished task
- An agent can pause a flow waiting for something and pick up at the exact step on the next message. A chatbot restarts the script.
- Rich replies
- A text-only reply is the tell. An agent sends carousels, option lists, forms, a catalogue, a cart and a payment request in the thread.
- A real back office
- A chatbot leaves you a transcript. An agent writes records — customers, orders, bookings, cases — that your team works in afterwards.
- Handover with context
- An agent opens a case, routes it, and attaches the conversation. A chatbot usually hands over a phone number.
- What a tap costs
- Ask whether a button press is billed as an AI message. Taps are the bulk of real traffic, and here they make no model call at all.
Why "it uses AI" is not the distinction
Plenty of chatbots call a model. That is not what separates them from an agent — a model call is a feature, not an architecture.
The distinction is whether the thing can carry STATE: whether it knows what is half-done, what the customer already said, what is in the cart, and what still has to happen before the job is finished. That is the difference between answering and running something.
Where a chatbot is genuinely the right answer
If what you need is an FAQ responder on a page with low traffic, an agent is more machinery than the problem deserves. A chatbot that answers twenty common questions and hands the rest to email is a perfectly good outcome.
The case for an agent starts when the conversation is meant to END in something — an order, a booking, a resolved case — rather than in an answer.
Where the honest line is on our side
- Messaging a customer outside WhatsApp’s 24-hour window needs an approved WhatsApp template, and that path is still coming. Any vendor — agent or chatbot — is under the same Meta rule, so ask them about the window specifically.
- We do not offer broadcast campaigns on WhatsApp today.
- "AI agent" is not a guarantee of quality. The six points above are checkable; the label is not. Ask for a demonstration against your own messages.
Questions people ask
- Is an AI agent just a chatbot with a better model?
- No. The model matters less than whether the system can carry state — what is half-done, what is in the cart, what still has to happen. A better model on a stateless script is still a script.
- Do I need an agent if I only answer questions?
- Probably not. If the conversation is meant to end in an answer rather than in an order, a booking or a resolved case, an agent is more machinery than the problem needs.
- How do I tell which one a vendor is selling me?
- Ask what happens when a customer comes back two days later mid-task, what the reply looks like beyond text, where the data lands, and whether a button press is billed as an AI message.
- Does either avoid WhatsApp’s 24-hour rule?
- Neither. It is Meta’s rule, not a vendor feature. Messaging outside the window needs an approved template regardless of who built the agent.
Keep reading
What a WhatsApp AI agent does · The business case for conversational agents · See one running
See which one you actually need.
Try it on your own website first — nothing is visible to a customer until you decide it is.
Get started free Browse the marketplace