Meet My Agents. Every One Started as an Annoyance.

In an earlier post I wrote that agents check my packing list, manage my travel, plan my meals, and draft my demos. A few people asked the fair follow-up question: show us.

So here they are. Not as a product pitch or an architecture lecture, but as what they actually are: a small team of digital helpers, and every single one of them was born from a real annoyance in my real life. That's the pattern I want you to notice, because it's the most useful thing in this post.

The packing agent, born in Stockholm

The origin story of this one is simple: I once arrived in Stockholm without a jacket. Stockholm. So I built an agent to make sure that never happens again.

Before a trip, it looks at where I'm going, for how long, and what the weather will be, and it checks my packing list against reality. Cold city, long trip, on stage or not. It doesn't pack my bag. It asks me the question I forgot to ask myself.

Is this world-changing technology? No. Did I have a jacket in every city since? Yes. That's the bar an agent has to clear: not impressive, just reliably better than my own memory.

The travel agent, born from a world tour

When your year contains dozens of flights, hotels, and time zones, your calendar stops being a schedule and becomes a puzzle. Confirmation mails everywhere, check-in times, the eternal question of what's actually booked and what I only think is booked.

So an agent keeps that puzzle solved. It reads the confirmations, keeps the schedule in one place, and answers the only question I actually have: what does my trip look like, and is anything missing? The information always existed. It was just spread across forty emails.

The agent's job is not knowledge. It's attention.

The kitchen agent, born on a busy Sunday

This one plans the week's menu and turns it into a shopping list, organized so I'm not walking back and forth through the supermarket like a lost tourist.

Note what it doesn't do: it doesn't cook. I wrote a whole post about why that line is exactly there. The planning and the shopping list are output, and output gets automated. The cooking is the point of my Sunday, so the cooking stays mine. This agent is the clearest example of drawing the line inside a single activity: automate the task, never the point.

The demo agents, born from deadlines

When I build a new demo, an agent reads the documentation and writes the first version. When something breaks, an agent gets the error before I do. This changed how I work more than anything else on this list.

But here's the nuance that matters: the agent does the typing, not the deciding. I choose what to build, I hit the real problems, and I don't put anything on stage that I can't explain down to the last decision. The rule from earlier in this series still stands: whoever is on stage must be able to answer the second question. Agents help me get there faster. They don't get there for me.

The editorial team, born from an experiment

The biggest one is not a single agent but a team: hamba.nl, a Dutch travel magazine where the editorial pipeline is run by agents. Research, drafts, editing steps, each agent with its own job, like a small newsroom where the staff happens to be digital.

I started it to learn what agents are actually like as colleagues, and the honest answer is: great at volume and consistency, terrible at taste and judgment. The pipeline produces, and a human decides. The interesting lessons from running it deserve their own post, but the short version fits here: the more agents you put to work, the more valuable the human at the end becomes.

What they have in common

Look back at the list. Not one of these agents started as an ambition. No "let's do something with AI." Every one of them started as an annoyance: a forgotten jacket, a chaotic calendar, a supermarket walk, a deadline, a curiosity about what this technology really is.

That's my honest advice if you want to start building agents.

Don't look for a use case. Look for an irritation.

The thing you grumble about every week is your first agent, and because the problem is real, you'll immediately know whether the agent works. No metrics discussion needed. You had a jacket or you didn't.

And one more pattern: every agent on this list has exactly one job. The packing agent doesn't book flights. The kitchen agent doesn't touch my calendar. Small agents with clear jobs beat one big agent that does everything vaguely. That's not just my experience at home. It's how the whole field is heading.

Build your own annoyance away

So that's my answer to "show us." Nothing on this list is spectacular, and that's the point. The spectacular stuff is on stage. This is just my life, running a little smoother, so there's more room for the parts I refuse to automate.

What's the annoyance you'd hand over first? Genuinely, tell me. The answers to that question are the best demo ideas I know.