A chatbot answers questions in a conversational interface. An AI agent takes multi-step action on a business's behalf, usually with some autonomy over what it does next. An AI decision system is scoped narrowly to one specific business decision and produces a recommendation a person can act on. All three get called "AI" in sales conversations. Only one of them tends to survive contact with a real Australian business past the pilot stage.
The terms are used almost interchangeably in the market right now, which makes buying decisions harder than they need to be. A vendor pitching a "chatbot" might mean a FAQ widget or a genuinely capable assistant. A vendor pitching an "agent" might mean something that reliably completes a defined task, or something that was built as an open-ended experiment with no clear boundary on what it is allowed to do. Knowing which one you are being sold matters more than the label on the pitch deck.
What is a chatbot, specifically?
A chatbot is a conversational front end. It answers questions, usually by matching an enquiry to a knowledge base or a script, and hands off to a person when it runs out of road. Its job is to reduce simple, repetitive enquiries, not to make decisions.
Chatbots are cheap to stand up and genuinely useful for narrow, high-volume questions: opening hours, order status, basic troubleshooting. The trouble starts when a chatbot is asked to do something it was never built for, matching a customer to the right one of 40 services, or assessing whether a situation needs escalation, because a conversational front end with no underlying decision logic will guess, and guessing at scale is expensive.
What is an AI agent?
An AI agent is software that can take a sequence of actions toward a goal with some degree of autonomy: querying systems, drafting responses, calling other tools, deciding what to do next without a person approving each step.
That autonomy is exactly why agent projects have a rough track record right now. Gartner's June 2025 research, based on a poll of more than 3,400 organisations actively investing in the technology, predicts more than 40% of agentic AI projects will be cancelled by the end of 2027. The stated reason is not that the technology fails. It is that most agentic AI projects are early-stage experiments driven by hype, deployed without a clear strategy or governance for what happens when the agent gets something wrong.
An agent without a tightly scoped decision underneath it is a demo, not a business system.
What is an AI decision system, and how is it different?
An AI decision system borrows the useful part of "agent," the ability to act on real inputs without a person doing every step by hand, and drops the open-ended part. It is scoped to one specific, repeated decision from the start: match this enquiry, assess this fault, triage this document. It has clear boundaries on what it can decide alone and what it must escalate.
That narrower scope is what makes it reliable enough to run in production. It also makes it possible to prove the system works, and to put a number on what it is worth, before committing to build it at scale.
Why does the distinction matter to what you buy?
Because the three things cost different amounts, carry different risks, and solve different problems. A chatbot is a reasonable, low-cost fix for repetitive enquiries. An open-ended agent is a research project with a real chance of being one of the 40% that gets cancelled before it delivers anything. An AI decision system is the one built to be judged on a specific business outcome from day one, prototyped against real inputs before any large commitment is made.
If a vendor cannot tell you, specifically, which single decision their system is built around and what happens when it is not confident, you are being sold a demo, not a decision system.
What should you ask before you buy?
Ask what specific decision the system makes, what happens when it is uncertain, and what evidence exists that it works on your business's real inputs rather than a generic demo. Any credible answer names one decision, not a general capability.
It is worth asking the question even of tools you already use. A generic AI subscription used for drafting emails or summarising documents is not doing anything wrong, but it is not a decision system either, and does not need to pretend to be one to earn its keep. The confusion only becomes expensive when a business pays agent-level or decision-system-level prices for something that was only ever built to be a chatbot, or commits to an open-ended agent build with no defined decision underneath it and no way to know in advance whether it will be one of the projects Gartner expects to be cancelled.
Qode builds AI decision systems, not open-ended agents or FAQ chatbots, starting with a working prototype against your real inputs before any production commitment. Book a Decision System Sprint fit call and find out which of the three you need, honestly, including if the answer is none of them yet.
