Eight things needed me today.

Earlier this month I built an attention planner for myself. An agent reads what came in, works out which things only I can decide, and books a block in my calendar for each one. Every decision gets a card with the question and a suggested answer. I write one line back, and the agent carries out the rest.

A week later I opened it and found four open cards. Every answer line was blank. Five focus blocks had come and gone.

Nothing was broken, and the cards were good. I had built one more place that wanted my attention, and it lost to the places that were louder. That week taught me more about planning a day than the week I spent building the thing.

Microsoft put a number on the louder places. Their 2025 Work Trend Index found that during core hours, people get interrupted about every two minutes. That adds up to around 275 interruptions a day, spread over roughly 117 emails and 153 chat messages. Almost half said their work feels chaotic and fragmented.

I doubt anyone reads that and feels lazy. They feel recognised. We still talk about the inflow as a discipline problem, something a better morning routine should fix. I have tried a lot of morning routines. None of them holds up against 275.

Herbert Simon saw this coming. In 1971 he wrote that “a wealth of information creates a poverty of attention.” A less quoted line from the same talk is more useful today. A new system only lightens everyone else’s load, he said, “if it absorbs more information previously received by others than it produces, that is, if it listens and thinks more than it speaks.”

That is the test for every agent we are about to point at our inboxes. An agent that reads 200 Slack messages and hands me 40 bullet points has failed it. I now have 40 things to read, written in a voice I trust less than the original. Researchers writing in HBR last year gave this kind of output a name, workslop. Around 41% of the people they surveyed had received some in the past month, and each instance took close to two hours to deal with.

So when I picture agents running parts of my day, the first thing I ask about is how much they swallow without coming back to me.

If agents are good at reading, the part left for me is the part I cannot hand over. Deciding what a situation actually is. Deciding what matters this week. Being the one accountable when it goes wrong. Taste. Whether I trust someone.

That changes what should reach me. Raw information should not. A decision should, written the way a good chief of staff would write it: the question, what the agent would do, and what happens if I say nothing. I can answer that in a line.

There is a limit here as well. People who approve things all day start approving without reading. So I keep very few of these open at once, and three is usually already too many.

Oliver Burkeman’s Four Thousand Weeks starts from something obvious that I still resist. You will never get through it all. The inflow has no end and you do. The useful skill is choosing what to neglect, on purpose, before it gets chosen for you.

His practical version is a short, closed list. A handful of things, and something only goes on when something comes off. Everything else waits somewhere out of sight.

Out of sight matters more than I expected. A 2018 study by Zhu, Yang and Hsee found that people choose urgent tasks over important ones even when the important task pays more. Knowing about the bias did not stop them. They called it the mere urgency effect. A red badge, then, is a thumb on the scale. If my day is going to be ordered by what matters, urgency has to be hidden by default. Pushing it down the list is not enough.

The finding I think about most comes from a 2019 field experiment by Fitz, Kushlev and colleagues. For two weeks, 237 people got their phone notifications in different ways. The group that got them in three batches a day was more attentive and productive, in a better mood, and felt more in control. Hourly batches did roughly nothing.

Then the part I did not expect. Turning notifications off completely made people more anxious. With no updates at all, they worried about what they were missing.

That explains my week with the planner better than I could. What I needed was updates I could predict. So the version I am setting up now reports three times a day, at times I know, and stays quiet in between.

Most of my switching happens between tools, about the same piece of work. A Figma comment, a Linear ticket and a Slack thread can be one conversation happening in three places, and I rebuild the context in my head every time I cross over.

Gloria Mark found that people who get interrupted make up for it by working faster, and pay for it in stress, frustration and a feeling of time pressure. Sophie Leroy showed that an unfinished task leaves “attention residue” that drags on whatever comes next. Every hop between tools leaves a little of that behind.

So I want the agent to group by topic. One card for each thing that is going on, with the thread, the ticket and the comment inside it. That removes more switching than any amount of batching.

I have run this as a list before. Lists grow. A list looks the same at 8 items and at 80, and every item carries the same weight.

On a canvas, position means something. In a 2010 study called “Space to Think,” people analysing documents on a large display used where things sat as a kind of memory. This corner held the sources they did not trust yet. That cluster was one argument. I want the same for a day. Roughly:

• A quiet count. What arrived, and how much of it agents handled without me.
• The few decisions that are actually mine, each with a suggested answer.
• One protected block for the thing I most want to move.
• What moved today, even if it was small.
• What I am waiting on from other people, so it stops living in my head.

The count is the piece I am most curious about. My guess is that seeing “212 arrived, 204 handled” will do something a clean inbox never did for me. It says the ignoring was done on purpose, by something I set up. Right now it feels like something I am getting away with.

Two pieces of research changed how I close the day.

Masicampo and Baumeister found that an unfinished goal keeps intruding on your thoughts, and that making a specific plan for it is enough to quiet it down. You can leave the thing unfinished. Writing the next step down is enough. So at the end of the day every open card gets one line: what happens next, and who does it.

The other is Kahneman’s peak-end work. We remember an experience mostly by its most intense moment and by how it ended. If the last thing I see at six is the pile I did not get to, that is the day I will remember. So the last thing on the board is what moved. Teresa Amabile’s diary research, around 12,000 entries from 238 people, found that progress on meaningful work, even small progress, was the thing most tied to a good day at work. I would rather end on that.

Two things.

Judgment is a muscle. If agents do all the reading and I only ever see a question with the answer already filled in, I am not sure I stay good at the reading that made my judgment worth asking for. I do not have an answer yet. For now I read some raw threads on purpose, the way you take the stairs.

The second comes from Byung-Chul Han’s The Burnout Society. His point is that burnout now comes less from a boss pushing us and more from us pushing ourselves, because we can. If agents give me back two hours, the easy thing is to fill them. Christina Maslach’s research names workload, control and reward among the main causes of burnout. A calmer board helps with control and makes progress easier to see. It does nothing for workload if I answer the extra time by taking on twice as much.

So this is the version I am living with for now. Agents that swallow most of it. A handful of decisions. One protected block. Updates three times a day. A day that ends on what moved. I will write again once I have lived with it for longer than a week.

• Herbert Simon, "Designing Organizations for an Information-Rich World" (1971)
• Microsoft Work Trend Index, "Breaking down the infinite workday" (2025)
• Niederhoffer et al., "AI-Generated Workslop Is Destroying Productivity," HBR (2025)
• Oliver Burkeman, Four Thousand Weeks (2021)
• Zhu, Yang and Hsee, "The Mere Urgency Effect," Journal of Consumer Research (2018)
• Fitz, Kushlev et al., "Batching smartphone notifications can improve well-being," Computers in Human Behavior (2019)
• Mark, Gudith and Klocke, "The Cost of Interrupted Work" (CHI 2008)
• Sophie Leroy, "Why is it so hard to do my work?" Organizational Behavior and Human Decision Processes (2009)
• Andrews, Endert and North, "Space to Think" (CHI 2010)
• Masicampo and Baumeister, "Consider it done!" Journal of Personality and Social Psychology (2011)
• Kahneman et al., "When More Pain Is Preferred to Less," Psychological Science (1993)
• Teresa Amabile and Steven Kramer, The Progress Principle (2011)
• Byung-Chul Han, The Burnout Society (2015)
• Leiter and Maslach, "Six areas of worklife" (1999)