What is an AI agent, in plain English?
An AI agent is software you give a goal to. It works out the steps itself, uses tools such as your email, files or business systems to carry them out, and checks its progress. A chatbot answers one question at a time. An agent keeps going until the job is done or it needs you.
What is the difference between an agent and a chatbot?
A chatbot waits for you to ask something, answers, and stops. An agent is given a job, such as "find out why this customer's order is missing", and then does the legwork itself: it decides which systems to look in, reads what it finds, and works towards an answer. You are not steering every step.
Three things make something an agent rather than a chatbot:
- A goal, not a question. You describe the outcome you want.
- Tools. It can look things up in, or act on, your email, files, spreadsheets or business systems, not only talk.
- A loop. It tries something, looks at what came back, and decides what to do next.
What does an AI agent do in a real business?
Here is one I have built. A customer reports a problem, say a missing order or an account that will not work. Instead of someone digging through systems, the agent:
- checks your systems, your guides and past cases;
- works out the cause;
- prepares the fix and the reply to the customer;
- stops and waits for a person to approve.
That investigation is the part that eats the time. For one person doing that job, the saving is typically about 15 hours a week (the figure on the examples page). I have also shipped agentic products that sit in front of enterprise customers.
Should an agent be allowed to act on its own?
For most small businesses, not at first. The safest pattern is that the agent investigates and drafts, and a person approves anything that leaves the building. That is how I build them: nothing sends without your OK. You get the speed of the agent doing the legwork and keep the judgement on the parts where a mistake would cost you a customer.
You can loosen that later, one task at a time, once you have seen how often it gets things right.
What should I check before using an AI agent?
- What can it reach? An agent can only see what you connect. Start with the least it needs. Can AI tools see my emails and files once connected? covers how permissions work.
- What can it change? Reading is low risk. Sending, deleting and paying are not. Keep those behind an approval.
- How will you know when it is wrong? Ask to see what it checked and why it reached its answer, not only the result.
- Who looks after it? Agents depend on other tools and keep working only while those stay the same. What happens when an automation breaks? explains what to plan for.
Questions people also ask
What is an AI sales agent, and should I trust one?
An AI sales agent is an agent whose job is sales tasks, such as researching prospects or drafting follow-ups. Trust it with the research and the drafts, and keep a person on anything sent to a prospect. The risk is not that it is useless, it is that a confident, wrong message goes out under your name.
What is an AI agent example?
An agent that investigates a customer problem is a good example. It checks your systems and past cases, finds the cause, and prepares the fix and the reply for a person to approve. The same pattern works for chasing missing paperwork from a new customer, or checking a weekly report against last week's.
What is the difference between an AI agent, an automation and a chatbot?
A chatbot answers questions. An automation follows fixed steps you set up in advance, the same way every time. An agent is given a goal and decides its own steps. Many useful systems combine them: a fixed automation to collect the information, and an agent to work out what it means. The full comparison is here.
Is agentic AI just hype?
Some of the claims are. The useful part is narrower: software that can work through a multi-step job using your tools, with a person approving the important steps. Judge it by one question: does it take a specific job off someone's desk and do it reliably?
See it working
Things I've built that do this, and the time they give back.
Customer problems investigated for you
Every problem a customer reports is checked against your systems, guides and past cases, with the cause found and the fix ready for your team to approve.
Typically 15 hrs a week per person
See how it works →