Chatbots answer questions.

Agents complete tasks.

That distinction sounds small. It isn’t.

What an Agent Actually Is

You don’t prompt an agent. You give it a goal.

It breaks the goal into steps. Executes them using tools — browsing, writing code, calling APIs, reading files. Adapts when something fails. Tries again.

The human sets the destination. The agent figures out how to get there.

“Research the top 10 competitors and summarize their pricing.” “Find all invoices in this folder, extract totals, output to spreadsheet.” “Monitor this page and alert me when the price drops.”

These aren’t responses. These are workflows. Running without you.

Why This Is Different From What Came Before

The chatbot era was reactive. You prompt, it responds. You’re always in the loop initiating every action.

Agents flip that.

Set the goal. Walk away. Come back to results.

That’s not an incremental improvement in the technology. That’s a different relationship with it entirely. The question shifts from “what can I ask AI?” to “what can I delegate to AI?”

Those are very different questions.

Where We Actually Are

Early days. Agents are impressive and unreliable in ways that matter.

They hallucinate steps. They loop. They do exactly what you said instead of what you meant — which are not always the same thing.

I’ve been building with this. An agent that reasons about game state in StarCraft II. One that coordinates between systems. One that makes decisions without being asked for each one.

It works. It also fails in ways that teach you more than the successes do.

The failures are the useful part right now.

The Part Worth Watching

An agent that reads your calendar, checks email, browses the web, writes code, sends messages — operating toward a goal you set — isn’t science fiction.

It exists today. It’s just not reliable enough yet for most people to trust with anything important.

Yet is doing a lot of work in that sentence.

The trajectory is clear. The reliability gap is closing faster than most people realize. When it closes enough — not perfect, just reliable enough — the way knowledge work gets done shifts in ways that make the chatbot era look like a warmup.

Watch this space.

Or better — build something in it.