AI agent vs chatbot: follow one inquiry through both
A chatbot answers the customer. An AI agent prices the job, books it, confirms it and sends the invoice. The fastest way to see which one you are being sold is to follow a single inquiry from start to finish.
Three different tools are sold under the word agent. A definition will not help you much in a sales demo. A test case will, so here is one.
It is Saturday night. Someone moving out of an apartment lands on a cleaning company's site and wants a move-out clean before Friday.
The same inquiry, three ways
With a chatbot
The customer asks what a move-out clean costs. The chatbot answers with a price range and says someone will be in touch. The conversation ends there. On Monday a person reads the transcript, calls back, and finds the customer booked someone else on Sunday.
The chatbot did its job. Its job was to answer.
With an automation
The customer fills in a form. A trigger fires, an email goes out that says thanks, and a row is added to a spreadsheet. If the customer asks a question nobody wrote a rule for, nothing happens. Somebody still has to quote, schedule and invoice.
With an AI agent
The customer picks the service and sees the price. They choose a time and it lands on the schedule. The day before, a confirmation text goes out. After the job, the invoice is sent, and if it is not paid it is followed up. One customer record carries all of it.
That last version is how Tidy Up Services runs in Jacksonville, where six of seven steps happen with nobody touching them.
Where the line sits
Gartner draws it plainly: use AI agents when decisions are needed, automation for routine workflows and assistants for simple retrieval.
| Tool | What it does | Where it stops |
|---|---|---|
| Chatbot or assistant | Answers a question when someone asks | It does not act in your systems, and it starts over when the chat closes |
| Automation | Runs the same steps every time a trigger fires | It stops at any case nobody wrote a rule for |
| AI agent | Takes a goal, decides the steps, acts and checks the result | It needs limits and review, which you should ask to see |
Most products called agents are not
Gartner has a name for this: agent washing. Chatbots, assistants and robotic process automation get rebranded as agents without real agentic capability. Gartner estimates that only about 130 of the thousands of vendors selling agentic AI are real.
It also predicts that more than 40% of agentic AI projects will be canceled by the end of 2027, because of escalating costs, unclear business value or inadequate risk controls. Buying the wrong category is one way to end up in that 40%.
When a chatbot or an automation is the right buy
An agent is not always the answer, and the cheaper tools are good at what they do.
A chatbot is the right buy when the job is answering the same questions over and over, and a person is ready to pick up everything else. If most of your inquiries are "are you open Sunday" and "do you service my area," a chatbot handles that well.
An automation is the right buy when the steps never change. A form comes in and a thank-you goes out. A payment clears and a receipt is sent. Fixed steps do not need a tool that makes decisions.
The trouble starts when either one is sold as something that will run the front office. Then the gap between the answer and the finished job lands back on you, usually on a Monday.
Four questions to ask in the demo
- Does it decide? Give it a request nobody planned for. An agent works out the steps. A chatbot replies with text, and an automation does nothing.
- Does it act? Ask what it changes on its own: a record, a booking, an invoice. Talking about the work is different from doing it.
- Does it remember? Ask what it knows tomorrow about today. An agent works from a shared record of your customers and their history.
- Does it check itself? Ask what happens before a customer sees its output. If the answer is nothing, a person is still doing the checking.
Ask the vendor to run your Saturday night inquiry live. Watching one request go from first message to invoice tells you more than any feature list.
Where this is heading
Gartner predicts that by 2028, 33% of enterprise software applications will include agentic AI, up from less than 1% in 2024, and that at least 15% of day-to-day work decisions will be made autonomously, up from none.
Xaedros is built as agents from the ground up. A Director routes every request to a standing team, every agent reads one shared record, and three independent reviewers score every customer-facing message before it goes out. See the architecture or what several agents do that one assistant cannot.