What they are, how they work, and why they matter now
Most software waits for you. You open it, click, it responds. An AI agent is different, you give it a job, not a click, and it goes and does it: reads what it needs, decides what to do, acts, and reports back. Less like a calculator, more like a capable colleague who works at machine speed.
That shift, from tool you operate to colleague you delegate to, is what makes this moment worth paying attention to.
Not everything with "AI" in the name is an agent. A chatbot that answers a question is a tool, useful, but passive. Something becomes an agent when it can do three things on its own:
It doesn't just talk, it reads files, browses the web, calls APIs, runs commands. Real actions, not just words.
ReadsWritesExecutesYou give it a goal, not a script. It works out the sequence of steps itself, and adjusts the plan as it learns more.
Goal, not stepsAdaptsBefore it calls a job done, it reviews what it produced, runs the tests, re-reads the output, and tries again if something's off.
Self-checksIteratesAn agent isn't a smarter model. It's a model given a loop, some tools, and permission to keep going until the job is done.
Strip away the buzzwords and every agent, simple or complex, runs the same loop:
Not a script of clicks. A brief, "fix the failing tests," not "open this file, change this line."
Reads what it needs, plans the steps, and uses its tools, files, a browser, an API, to actually do the work.
Reviews its own output, catches mistakes, and tries again, before calling the job done.
This loop repeats until the job is genuinely finished, or until it needs you.
This is exactly what's happening when you hand Claude Code, Codex, or Cursor a real brief, the same loop, whether the job takes ten seconds or ten minutes.
Every career is becoming an AI career. Whatever you do next, you'll do it alongside agents like this. The people trusted with bigger work will be the ones who can brief an agent clearly, steer it when it drifts, and stand behind the result, not the ones who could build one from scratch.
Creativity & delegation, seeing what's worth building, and turning it into a brief an agent can run with.
Steering & judgment, spotting the moment it drifts, pulling the work back on course, and never shipping what you can't explain.
The reasoning model is already built. The infrastructure is already managed. What's left to learn is judgement, what to hand off, what to check, and when to stop and ask. That's a practice problem, not a coding problem. It's exactly what the briefs on this site are for.
A 1-hour hands-on session where you explore the magic, possibilities and limitations of AI Agents.
Programme
What is an AI agent? Core concepts, Harness, Skills, Plugins, MCP, in plain English. A live demo shows Claude CoWork autonomously navigating a computer and solving a real problem without human input.
Set up and launch your first agent using Claude CoWork. Give it a real-world task, read a folder of messy documents, organise the information, and write a clean summary, all on its own.
A guided tour of where to go after today, teaching your agent a repeatable skill, connecting it to real tools via MCP, and Claude Code for work that lives in files on your computer.
Ask anything, then leave with a Prompt Cheat Sheet and a clear path to your first brief on the platform.
What you need on the day