Probably not. Most business jobs that people want an AI agent for are better served by a fixed workflow or a plain AI assistant. An agent earns its place when the steps can't be planned in advance, and that's rarer than the marketing suggests.
The word "agent" is currently stuck onto all sorts of things, including chatbots and ordinary automations that have simply been renamed. That makes it hard to know what you're being offered, and harder to know whether you need it. The test below cuts through most of that.
An AI agent is software that is given a goal and decides for itself what steps to take to get there, using tools such as your email, calendar or a web browser along the way. You tell it the outcome you want. It works out the route.
The usual way to separate them is by who decides what happens next.
Plenty of real systems mix these. A chatbot at the front, with a fixed workflow doing the work behind it, is common and often the sensible design.
A simple workflow is better whenever you can write the steps down in advance. Sending a confirmation when a booking comes in, chasing an unpaid invoice after a set number of days, or filing an enquiry into the right folder are all workflows. They're cheaper to run, easier to test and much easier to explain to a customer when something goes wrong.
If you can describe the job as a list of steps, you almost certainly don't need an agent. This is also where a spreadsheet, a template or a single well-written prompt often wins.
Before you spend anything, we'll tell you honestly whether it's an agent, a workflow or something simpler that would do the job.
GET A SECOND OPINIONAn AI agent is worth considering when the work is varied, the path can't be mapped out beforehand, and a mistake is cheap to catch and fix. Research tasks that pull from many sources, or sorting a messy mix of incoming requests, are the typical examples.
Even then, the sensible way in is small. Give it one narrow job, limit what it can see and do, and keep a person checking the results until you trust it.
The ICO, which regulates data protection in the UK, has published a report on agentic AI that is a useful checklist even if you never read the rest. Its main point is that an agent doesn't take responsibility away from you. The business using it stays responsible for what it does with personal data.
On its own, a chat window is mostly a chatbot: you ask, it answers. Some products add agent features that can carry out several steps or use other tools for you. The label depends on whether the software decides its own steps, not on the brand.
The main ones are that an agent can misunderstand an instruction, pass on an error to later steps, access more information than it needs, or act without anyone noticing for a while. All of these are reduced by narrow jobs, tight permissions and a person checking the output.
You are. The ICO is clear that AI agents are not legal entities and that the organisation using one stays responsible for how personal information is used. Blaming the software doesn't change that.
It varies a lot. The cost depends mostly on how much the agent has to connect to, how much needs checking, and how much running and maintenance it needs afterwards. The honest first step is working out whether you need one at all.
Tell us what's slowing you down.
We'll tell you honestly if we can help.