"We want an AI chatbot." We hear it constantly, and about half the time, what the business actually needs is an agent. The words get used interchangeably, but the gap between them is the gap between answering a question and getting a job done.
A chatbot answers
A chatbot holds a conversation. It responds to questions, points people to information, qualifies a lead by asking a few things, and hands off to a human when it is out of its depth. Trained well on your business, it is genuinely useful: it covers your support inbox around the clock, catches leads while you sleep, and stops you losing people who message at 11pm.
But notice what it does not do. It talks. It does not reach into your other tools and change anything. Ask a chatbot to actually reschedule the booking, update the record, and send the confirmation, and a basic one just tells you how to do those things yourself.
An agent acts
An agent is the one that does them. It takes the same conversation and then takes action across your systems: checks live availability, moves the booking, updates the CRM, sends the confirmation, and flags anything it could not resolve. It does not just describe the outcome, it produces it.
A chatbot is a conversation. An agent is a colleague.
That is the whole distinction, and it decides everything about cost, build time, and what you should expect. An agent needs permission to act inside your tools, a clear sense of what it is and is not allowed to do on its own, and a reliable handoff to a human at the edges. It is more to build. It is also worth far more, because it removes the work, not just the questions about the work.
Why the mix-up gets expensive
Businesses burn their first AI budget in two directions. Some pay for an agent when a chatbot would have done, building complex tool access for something that only ever needed to answer FAQs. Others buy a chatbot expecting it to run their operations, then feel let down when it politely explains the process instead of doing it.
Both are avoidable with one question, asked before anyone builds anything.
The question that sorts it
After the conversation, does something need to happen in another system?
If the goal is to inform, qualify, or route a person, you want a chatbot. If the goal is to complete a task, change a record, book a slot, trigger a workflow, you want an agent. Support desk that answers questions and passes hot leads on: chatbot. System that takes a request and fulfils it end to end: agent.
Plenty of businesses want both, and the natural shape is a chatbot front that escalates into agent actions when the request calls for it. That is fine. The point is to name which one each part of the job is, so you build the right thing and pay for exactly what the problem needs.
Where to start
If you are new to this, a well-built chatbot on your busiest channel is a low-risk, high-visibility first step. It earns trust, and it shows you exactly which conversations keep ending in "now a human has to go do the thing." Those conversations are your agent shortlist. If you are still deciding where to point a first build, start with where automation pays off first.
Whatever we build, you own it, with full documentation and handover on request. Plenty of teams then keep us on to grow it from a chatbot into an agent, as their volume and their confidence build.
Chatbot or agent for your business?
Book a call and we will tell you which one your problem actually needs, and why.
Book a 30-min call →