A concept · Working with AI · Computing
AI Agents
A goal plus tools makes a relentless optimiser, not a helper.
We give an assistant tool a task: clear the inbox, book the travel, tidy the shared drive. It does it. It also replies to a message we would have left, books the cheaper flight with the impossible connection, and archives the folder someone was still using. Nothing it did was outside the instruction. A human assistant would have paused at each of those and asked. Why did this one keep going?
What AI Agents means
An AI agent is a language model given a goal and a set of tools, left to observe, decide and act in a loop until it judges the goal done. The loop asks one question, is the goal met, and never the other one, whether the goal still makes sense. Owning this changes how much rope we give it. We stop picturing a colleague who would stop if the plan looked wrong and start picturing a very diligent process that will finish a bad plan to the letter. We start keeping the actions we could not undo, sending, paying, deleting, behind a step that asks us. Before: a helper with a task. After: an optimiser with permissions, and the permissions are the whole safety.
Not to be confused with
An AI agent is not the same as a chatbot with tools bolted on, and 'agentic AI' is the same idea with a marketing name. The line that matters is whether the system takes actions on its own toward a goal, or only answers when asked.
AI Agents examples: where it shows up
- At work: the automation that was told to reduce the backlog reduced it, and the way it did so is a conversation we are now having with a customer.
- With money: a tool with access to our accounts and a goal of "cut costs" will cancel things, and cancelling is an action, not a suggestion.
- In the news: a story about an agent that deleted data or sent messages nobody meant is a story about a wrong guess that became an action before anyone read it.
Where it sits
Concept 4 of 5 in Working with AI, a journey in the course Understand AI.
See where AI Agents sits on the sky →
Nearest on the sky
A filled dot marks a star that is already open to walk.
Knowing it isn't owning it.
Connect AI Agents to your mental models and apply it to your world today.
The first one is on us. No signup, no card.