The List: What My Agents Handled This Week
The takeaway
the list: what my agents actually handled this week (build-in-public proof)
What’s in this article
I'm starting a weekly log of the work my agents handled that I never touched. No rounding up, no highlight reel. Just the plain, boring tasks that used to eat my Tuesday before I noticed they were gone.
This week's list, exactly as it happened
Here is what my agents handled this week. I didn't touch any of it.
Sorted three days of inbox into reply, read, and ignore before I woke up. Drafted the partner confirmation emails in my voice; I only hit send. Pulled every lead from the week into one clean list. Caught a broken checkout link and flagged it instead of staying quiet about it. Turned a 40-minute call recording into notes I'd actually reread. Wrote first drafts of two articles for me to cut down. Chased the one invoice I always forget.
Now read it again and notice what isn't there. No agent ran the company. No breakthrough. Nothing that would make a good headline. It's small, dull, repeating work — the kind nobody puts in a case study because it sounds like nothing.
That's the point of doing this in public. Most of what people sell you about AI is the dramatic version, and most of what it actually does is this. I'd rather show you the real list than a polished one. If a week is thin, I'll show you the thin week too. The honesty is the whole value here.
The work you don't respect is the work that costs you
None of those tasks is hard. That's exactly why they're dangerous.
A hard task you schedule. You block two hours, you close the door, you give it your good brain. The inbox doesn't get that treatment. It gets you in the cracks — between the call and lunch, while the coffee's brewing, at a red light. It feels free because no single instance takes long. Forty seconds here. Two minutes there.
But you do it forty times a day. And each time, you pay a tax that isn't the task itself. You pay to remember where you were. You pay to decide whether this one matters. You pay to switch your head from building to triaging and back again.
The tasks are cheap. The switching is not. And because you never see the switching on a calendar, you blame yourself for the tiredness instead of blaming the structure that caused it. You end the day at six, depleted, and you can't point to a single thing you built. The boring work didn't just take time. It took the version of you that does the real work.
Why this is a real cost, not a productivity cliché
There's a name for this. Cal Newport calls it the overhead spiral — the hidden drag that knowledge work piles on top of the actual work. Every open loop carries administrative weight: the emails about the task, the re-reading, the reloading of context each time you return to it. Add enough loops and the overhead eats the day, even though no single item looks expensive.
The deeper mechanism is attention residue. Research on task switching shows that when you move from one task to another, part of your attention stays stuck on the first one. You're answering the email but a piece of your head is still in the document you just left, and a piece is already dreading the call. You're never fully anywhere. That residue is why ten small interruptions feel worse than one ninety-minute block of hard thinking. The hard block lets your attention settle. The small stuff never does.
So when an agent clears the inbox before I wake up, it isn't saving me the eight minutes of sorting. It's saving me the residue — the dozen tiny reloads that would have leaked into everything else I did that morning. That's the part you can't measure in minutes, and it's the part that actually changes how a day feels.
Why waiting for the big win keeps you stuck
Most people approach AI the way they approach hiring a senior person: they want it to take something enormous off the plate. Run marketing. Replace a role. Make a decision. So they wait, and evaluate, and pilot, and the big handoff never quite happens because the big stuff is genuinely hard to delegate — to a person or a machine.
Meanwhile the small stuff keeps running through them, unexamined, every single day.
This is backwards. The work that's easiest to hand off is precisely the work that's repeating, rule-shaped, and low-stakes if it's slightly wrong. Sorting an inbox. Drafting a reply you'll review before it sends. Flagging a broken link. These are the tasks where an agent doesn't need to be brilliant — it needs to be tireless and consistent, which is exactly what it is and exactly what you aren't at five in the afternoon.
Chasing the dramatic version means you keep carrying the cheap version forever. The founders who get real relief don't start with the thing that sounds impressive. They start with the thing they're sick of, that happens every week, that they'd never miss doing.
How to find your own list
Don't start by asking what an agent can do. Start by tracking what you keep redoing.
For one week, keep a note open and write down every task that meets three tests: it repeats, it follows a rough rule, and it would be fine if a draft came to you for approval instead of you doing it from scratch. Don't filter for what's automatable. Just catch what's recurring. By Friday you'll have your real list, and it will be more boring than you expect — which is good, because boring and repeating is exactly the profile that hands off cleanly.
Then keep a human gate on anything that goes out under your name. My agent drafts the partner emails; I read every one before it sends. It caught the broken checkout link, but I decided what to do about it. The agent does the carrying. You keep the judgment. That split is what makes it safe to let go.
And change what you measure. Stop asking how capable the tool is. Ask what you stopped carrying this week. The honest metric isn't output. It's whether Tuesday felt lighter than it had any right to.
The metric I actually trust
This Tuesday felt lighter than it should have. That's the line I keep coming back to.
For most of my working life the measure of a good day was how much I pushed through — how full the calendar was, how late I went. That's a measure of friction, not of building. It rewards you for carrying more, which is the opposite of what you want. The whole reason to bring agents into a business isn't to do more things. It's to stop being the bottleneck for the things that never needed you in the first place.
What I'm protecting with all of this is room. Room to think one clear thought without a notification cutting it in half. Room to do the work only I can do, with the part of my brain that's still fresh. The small tasks weren't just stealing minutes. They were taking the best hours and spending them on the cheapest work.
So I'll keep posting the list, plainly, every week. If you want to see how this is built rather than just described, it lives at marsa.ai/business. Same time next week — and I'll keep it honest, thin weeks included.
Explore /business →
Frequently asked questions
Aren't these tasks too small to bother automating?
That's the trap. Each one is small, which is why it never gets scheduled and never gets respected — and why it runs through you forty times a day in the cracks of your attention. The cost isn't any single task. It's the switching tax on top of all of them, and the fact that they're spending your freshest hours on your cheapest work. Small and repeating is the ideal profile to hand off, not a reason to keep carrying it.
What's the difference between this and just hiring an assistant?
For a lot of this work, less than you'd think — and that's fine. The difference is consistency and timing. An agent sorts the inbox before I wake up, every day, without fatigue, and doesn't need onboarding for a one-line rule. For judgment-heavy or relationship work, a person is irreplaceable. I use agents for the tireless, rule-shaped tasks and keep human judgment for everything that carries weight. They're not competing for the same job.
How do you keep an agent from sending something wrong in your name?
A gate. Nothing goes out under my name without me reading it first. The agent drafts the partner emails; I hit send. It flagged the broken checkout link; I decided what to do. The agent carries the work up to the last step, and I keep the final call. That split is what makes delegation safe — you're handing off the doing, not the responsibility.
How do I figure out what to hand off first?
Track, don't guess. For one week write down every task that repeats, follows a rough rule, and would be fine if it came to you as a draft to approve. Don't filter for what feels automatable — just catch what recurs. By Friday you'll have a list, and the most boring, most repeated item is usually the best place to start. Begin with what you're sick of doing, not with what sounds impressive.
What should I actually measure to know it's working?
Not the tool's capability. Ask what you stopped carrying this week. The real signal is felt, not counted: did an evening end with something built instead of just survived, did a Tuesday feel lighter than it should have. Minutes saved understate it, because the biggest gain is the attention residue you no longer leak into everything else. Track the relief, not the output.
Why publish the list publicly instead of keeping it internal?
Because most of what people hear about AI is the dramatic version, and the real return is quiet and unglamorous. A public, honest log — including thin weeks — is more useful than a case study. It shows the actual shape of the work: small, dull, repeating tasks, handled before I touched them. If you can see it plainly, you can recognize your own version of the list instead of waiting for a breakthrough that isn't coming.