Three words get used as if they mean the same thing, and they do not. The difference matters because it determines what you can hand over and what you cannot.
A chatbot answers
A chatbot waits for someone to type, then replies. That is the whole shape of it. Good ones are genuinely useful — answering “what are your hours” a hundred times a day is real work that nobody enjoys.
But a chatbot is passive. It does not know anything happened unless a person tells it. Nothing in your business changes because a chatbot exists; the phone just rings less.
An automation follows rules
An automation runs a sequence you defined in advance. When a form is submitted, add a row to a spreadsheet and send an email. Zapier and Make are excellent at this and cost almost nothing.
The limit is that an automation has no judgement. Every decision inside it was made by you, in advance, in a rules editor. Take a real case: a customer is 40 days late on an invoice.
An automation can send them a payment reminder. It will send the payment reminder — the one you wrote in the template, identical for every customer. Your best client of eleven years and the guy who is always late get the same words.
That is fine until it is not. Templates are how relationships get damaged by software.
An agent decides
An agent looks at the specific situation, reasons about it, and produces something that did not exist before.
Same overdue invoice. The agent sees: this client, this amount, 40 days, their payment history, whether you have chased them before, how they responded last time, and how much of your quarter depends on them. Then it writes a reminder for that situation — firmer for a repeat offender, gentler for a good customer having a hard quarter, more formal if the amount is large.
The message it writes at 9am Tuesday has never existed before and will never exist again in that form. That is the difference. Not “smarter automation” — a different category of thing.
The test
Here is how you tell them apart in a sales meeting:
Ask the vendor: “If two of my customers were in almost the same situation but not quite, would your system do exactly the same thing for both?”
If the answer is yes, it is an automation. That may be all you need — automations are cheap, reliable and boring in a good way. But do not pay agent prices for one.
If the answer is no, ask the follow-up: “What stops it doing something stupid?” The honest answer involves a human approving actions before they leave the building, at least until you have watched the thing be right enough times to trust it.
Where each one belongs
Use a chatbot for questions that have one correct answer and get asked constantly.
Use an automation for sequences with no judgement in them: move this file there, notify that channel, create a record when a form comes in. Most businesses should have more of these than they do, and they are nearly free.
Use an agent where the work requires reading a situation: writing a quote, deciding a reminder’s tone, diagnosing why output dropped, working out which of forty things that happened today actually needs you.
Most real deployments are all three at once. The automation notices and routes, the agent thinks and writes, the human approves. Anyone selling you only one of the three is selling you their product rather than a solution.
Orbys is the agent layer, with the automation and approval plumbing around it. If you want to see the difference on your own data, book thirty minutes.