Aviana Le · Creator + Founder · Building in publicInstagram @avianaleMogul Mastery waitlist

Mogul Maxxing  /  Blog  /  AI Second Brain

02 · AI Second Brain

A 60-minute work window plan, with AI doing the setup

An hour is enough to move a business forward if the hour starts with the work rather than with figuring out what the work is. Here is a minute-by-minute plan for one focused work window, where AI handles the briefing, the parking and the notes, and you keep the thinking. Plus what the research says, and where it's less settled than productivity advice suggests.

By Aviana LeNovember 202622 min read
Woman founder at a cream linen desk in soft morning light with an open laptop, a small brass timer and a chocolate leather notebook, ready to start a 60-minute work window
One hour, one output. The setup happens before the timer starts, so the hour can go to the work.

A 60-minute work window plan splits one hour into a short setup, a single protected block of work on one chosen output, and a short close-down that makes the next window easier to start. AI is useful at both ends: it can brief you on where the project stands in under five minutes, hold the stray thoughts that try to pull you away mid-block, and turn your rough end-of-window notes into a clean handoff. The middle stays yours. The plan I'm testing looks like this: 5 minutes of briefing, 5 minutes to name the one output, 40 minutes of single-task work, and 10 minutes to log decisions and set up the next start.

The short answer

Minutes 0–5: ask your AI second brain for a brief on the project. 5–10: you choose one output and write an if-then plan for the most likely interruption. 10–50: work on that output only; when a new thought shows up, park it with the AI in one line and return. 50–55: log what you decided and why, in your words. 55–60: have the AI draft a next-window brief from your notes, check it, and close. Research on task switching, planning and AI at work supports the shape of this plan. It doesn't prove any particular timing is optimal, so treat the minute marks as a starting point.

This article picks up a thread that runs through the series. The decision fatigue article described the pattern of sitting down with sixty minutes and losing the first twenty to rebuilding context. How to decide what to work on first covered choosing the task. And what an AI second brain should do for a founder gave the assistant a job description. This piece puts those together inside a single hour, because that's the unit most founders with other commitments actually have.

I'm testing this plan in the build, alongside a first offer I'm validating in public. I'm not going to tell you it doubled my output, because I haven't measured that and I wouldn't trust the number if I had one week of data. What I can share is the structure, the research behind each part, and the places where I think the popular advice overreaches.

Why a single hour goes missing

When an hour of work produces very little, it's tempting to call it a discipline problem. More often it's a startup cost problem. Before any real work happens, you have to remember where you were, decide what matters now, and push aside whatever you were doing before you sat down. If the window is short, those costs take a large share of it.

Three findings explain most of that cost, and each one points to a specific part of the plan.

Attention residue: the last task follows you in

Sophie Leroy, then at the University of Minnesota, published a set of experiments in 2009 in Organizational Behavior and Human Decision Processes on what she called attention residue. When people switched from one task to another, part of their attention tended to stay with the first task, especially when it was left unfinished, and their performance on the next task suffered. Finishing the first task wasn't always enough on its own; residue was also higher when people had been working on the first task under low time pressure.

For a founder, the first task is often not work at all. It's the school run, a client call, a shift at another job or a scroll through messages. The plan below uses the first five minutes partly to give that residue somewhere to go, by writing down what's still open before you start.

Switching and interruptions cost more than they seem to

Research on task switching has measured these costs for decades. Joshua Rubinstein, David Meyer and Jeffrey Evans reported in 2001, in the Journal of Experimental Psychology: Human Perception and Performance, that switching between tasks consistently took longer than repeating the same task, and that the cost grew with task complexity. Those were lab tasks measured in fractions of a second, so the exact numbers don't map onto a workday. The direction does.

At the scale of real work, Gloria Mark, Daniela Gudith and Ulrich Klocke studied interruptions in a 2008 paper at the CHI conference, The cost of interrupted work: more speed and stress. Interrupted participants completed their tasks in less time without a loss in quality, but reported significantly more stress, frustration, time pressure and effort. That's a more nuanced picture than "interruptions destroy productivity". People compensate by working faster, and they pay for it in strain.

You'll also see the claim that it takes "23 minutes" to get back on task after an interruption. That figure comes from Mark's observational research on office workers and has been widely repeated in interviews and articles, often stretched well beyond what was measured. I'd treat it as a sign that resumption can be slow, not as a law of attention. The practical point holds either way: in a 60-minute window, every interruption you can prevent or defer is worth preventing.

The planning fallacy: an hour holds less than you think

Roger Buehler, Dale Griffin and Michael Ross documented the planning fallacy in a 1994 paper in the Journal of Personality and Social Psychology. In one study, students predicted how long their honors thesis would take; their average prediction was about 34 days and the average actual time was about 55, and only around 30% finished within the time they predicted. People tended to focus on how the plan would go rather than on how similar plans had gone before.

In a single hour, the planning fallacy looks like a to-do list with six items on it. The plan below picks one output on purpose. If you finish early, you can always pick a second one. You can't get back the minutes spent switching between six.

A short work window isn't too short to matter. It's too short to spend deciding.

What the research says about AI inside a work session

The evidence on AI and productivity is growing quickly, and it isn't all pointing the same way. That's exactly why the plan gives AI specific jobs at the edges of the hour rather than a blanket role in the middle.

Large gains on well-defined writing tasks

Shakked Noy and Whitney Zhang, economists at MIT, ran an experiment with 453 college-educated professionals, published in Science in 2023. Participants did occupation-specific writing tasks, such as press releases, short reports and analysis plans. Those given access to ChatGPT took about 40% less time, and the quality of their output, rated by evaluators in the same professions, rose by about 18%. The gains were largest for participants whose first task was rated weaker.

Gains inside the frontier, losses outside it

The Harvard Business School field experiment led by Fabrizio Dell'Acqua with Boston Consulting Group, which I covered in more detail in the AI second brain for founders article, found something similar for consultants on tasks inside what the authors called the AI's "jagged technological frontier". On a task chosen to fall outside it, consultants using AI were less likely to reach the correct answer than those working without it. The lesson is that the same tool can help or hurt depending on the task.

And a result that cuts the other way

In July 2025, the research nonprofit METR published a randomized trial with 16 experienced open-source developers working on 246 real tasks in large codebases they already knew well. When the developers were allowed to use AI tools, they took about 19% longer to finish tasks than when they weren't. Before the study, they expected AI to speed them up by about 24%, and even afterwards they estimated it had made them about 20% faster.

The authors were careful about what this does and doesn't show. It was a small group of highly experienced people working on complex, familiar projects, and the tools have changed since. It doesn't mean AI slows everyone down. It does show something every founder should take seriously: how fast AI makes you feel and how fast it actually makes you can be different things. In a 60-minute window, an hour of prompting, waiting and reviewing can feel productive while the actual output barely moves.

Put together, these studies suggest using AI where the task is well defined and easy to check, such as summarizing your own notes, drafting a first version or tidying captured thoughts, and being much more deliberate about letting it into deep, unfamiliar or high-judgment work.

StudyWho and whatFindingWhat it means for one hour
Noy & Zhang (2023)453 professionals, writing tasksAbout 40% less time, about 18% higher rated quality with ChatGPTDrafting and tidying are good AI jobs
Dell'Acqua et al. (2023)758 BCG consultants, realistic tasksBetter and faster inside the frontier, less accurate outside itChoose the AI's jobs by task, not by habit
METR (2025)16 experienced developers, 246 real tasksAbout 19% slower with AI, while believing they were fasterFeeling productive isn't evidence; protect the deep block
Leroy (2009)Lab experiments on task switchingUnfinished prior tasks leave attention residueSpend the first minutes clearing what's open

The 60-minute work window plan, minute by minute

Here's the full structure. The minute marks are a default, not a rule. If your briefing takes three minutes, start working at minute eight. What matters is the order: context, then commitment, then work, then handoff.

MinutesPhaseYour jobThe AI's job
0–5BriefRead the brief, note what's still in your head from beforeSummarize the project: goal, last decision, next step, open questions
5–10CommitName one output and write one if-then planNothing, or check the output is small enough for 40 minutes
10–50MakeWork on the one output onlyHold parked thoughts; draft only if drafting is the output
50–55LogWrite what you decided and why, in a sentence or two eachFile the decisions with the date
55–60Hand offCheck the next-window brief and closeDraft the next-window brief from your notes

Minutes 0–5: the brief

Start by asking your AI second brain for a short brief on the project you're about to work on. If you followed the setup in how to use AI as a second brain, it already has your notes, your decision log and your open loops to draw from. A prompt like this works:

Briefing prompt

"Brief me on [project] in five lines, using only my notes: the goal, the last decision I made and why, the next step I wrote down, anything still unresolved, and anything I parked for this window. Show where each line came from. If my notes don't say, write 'not in notes'."

Check the brief against the source lines it points to. That takes a minute and catches the occasional confident mistake. Then, before you move on, write down anything still hanging over from whatever you were doing before you sat down. One line each is enough. This is the attention residue step: you aren't solving those things, just moving them out of your head and onto a list.

Minutes 5–10: commit to one output

This is the step I think matters most, and it's the one AI shouldn't do for you. Pick a single output you can realistically finish or meaningfully advance in about 40 minutes. Not "work on the sales page" but "write the first draft of the sales page's opening section". Not "think about pricing" but "write three pricing options with one sentence of reasoning each". If you're unsure which output to pick, the decision rules in how to decide what to work on first are built for exactly this moment.

Then write one if-then plan for the interruption most likely to derail you. Peter Gollwitzer, who developed the concept of implementation intentions, and Paschal Sheeran reviewed 94 independent tests of the idea in a 2006 meta-analysis in Advances in Experimental Social Psychology. They found a medium-to-large average effect on reaching goals compared with simply intending to reach them. The plans work by linking a specific situation to a specific response, so the decision is made in advance.

For a work window, that might be: "If I think of something else I need to do, then I'll type it into the parking note in one line and go straight back." Or: "If I get stuck on the headline for more than five minutes, then I'll write the worst acceptable version and move on." One good if-then plan beats a list of rules you won't remember mid-block.

Minutes 10–50: the make block

Forty minutes on one output. Close everything that isn't needed for it. If the AI isn't part of the output, close that too, apart from the one place you use for parking.

The AI's job here is mostly to be a parking lot. When a new idea, worry or task shows up, give it one line, such as "park: reply to the supplier about delivery dates", and return to the work. You don't need to organize it; the AI can label and sort parked items at the end. The point is that writing it down lets you stop holding it.

That last claim deserves a caveat. E. J. Masicampo and Roy Baumeister found in 2011 lab studies that making a specific plan for an unfinished goal reduced its intrusive effects on a later task. That's encouraging. But the broader "Zeigarnik effect", the popular idea that unfinished tasks are reliably remembered better than finished ones, has a weaker evidence base than its reputation suggests. A 2025 meta-analysis by Riccardo Ghibellini and Beat Meier found the evidence for a general memory advantage for unfinished tasks to be inconsistent, while the tendency to go back to and resume interrupted tasks held up better. So I'd frame parking as a sensible, low-cost habit with some supporting evidence, rather than as a proven fix for a distracted mind.

If drafting is the output, AI can be part of the block. Even then, I'm testing a rule from the earlier article: write the rough version or the outline yourself first, then ask for critique or a tighter version. It keeps your thinking in the work and makes the AI's contribution easier to judge. Given the METR finding, I'm also watching for the moment when I'm prompting and re-prompting instead of writing. If I've asked for the same thing three times, I stop and write it myself.

What about breaks?

A 40-minute block is close to the length many people can hold focus comfortably, but I don't want to dress that up as a scientific optimum. The popular claim that people work in natural 90-minute cycles traces back to Nathaniel Kleitman's idea of a basic rest-activity cycle, and the evidence that it governs daytime work performance is thin. The Pomodoro technique, which uses 25-minute blocks, also has limited direct research behind it.

One study is worth knowing about. Felicitas Biwer and colleagues at Maastricht University compared fixed "Pomodoro" breaks with self-chosen breaks among university students, published in 2023 in the British Journal of Educational Psychology. Students who took breaks whenever they chose had longer study sessions and longer breaks, and reported more fatigue and distractedness and lower concentration and motivation than students on fixed schedules. Task completion was similar in both groups, so the fixed schedules got the same work done in less time. It was a study of students, not founders, but it suggests that deciding your break point in advance is at least as good as relying on how you feel. In a single hour, that break point is simply the end of the make block.

Overhead view of a cream desk with an open notebook showing a hand-drawn hour divided into blocks, a small brass timer, a closed laptop and a ceramic cup of coffee
Context, commitment, work, handoff. The order matters more than the exact minute marks.

Minutes 50–55: log decisions

Stop working with ten minutes left, even if you're in the middle of something. This feels wrong the first few times. It's what makes the next window start fast.

Write down any decision you made during the block, in one or two sentences: what you chose, what else you considered and why you chose it. "Opening section leads with the problem, not with my story, because the three people I spoke to described the problem first." Then ask the AI to file the decisions with the date. As the AI second brain article argued, the reasoning should be in your words. A decision log the AI wrote is a record of reasoning you may not have done.

Minutes 55–60: hand off to your next window

Finally, write a rough note on where you stopped: the exact next step, anything half-finished, and anything you'd want to know when you sit down again. Then ask the AI to turn your note and the parked items into a next-window brief.

Handoff prompt

"Here's where I stopped: [rough note]. Using this note, today's decisions and my parked items, draft the brief I should read at the start of my next window on this project: goal, where I stopped, exact next step, open questions. Then list the parked items that belong to other projects, labeled, so I can file them. Don't add anything that isn't in my notes."

Read it, fix anything that's off, and close. The next time you sit down, minutes 0–5 are already half done. This is the same idea as a shutdown ritual, described in the founder operating system article, shrunk to fit inside a single hour.

Mogul Mastery · founding waitlist

Make every work window start with the work

Mogul Mastery is the implementation program I'm building in public for founders with limited work windows: decision rules, operating rhythms and AI used for context and decision support, set up around your own business. It's in development and shaped by the people on the founding waitlist.

Adapting the plan to the kind of hour you have

Not every window is the same. Here's how I'd adjust the plan for the most common situations, keeping the order and changing the proportions.

Kind of windowAdjustmentWhere AI helps most
Deep making (writing, designing, building)Keep the full 40-minute block; keep AI out of the middle unless drafting is the outputBriefing and handoff
Admin and repliesBatch similar items into the block; set a hard stopDrafting routine replies you then edit
Customer conversationsUse the block for the call; use the log phase to capture what was saidSummarizing your notes into quotes and themes
Thinking and strategyShorten the block to 30 minutes and lengthen the log to 15Laying out options you've written, not choosing between them
Learning a new skillYou attempt first, then ask for critiqueQuestions and feedback, not answers
A window that starts lateCut the make block, never the handoffA faster brief from the last handoff

If you run more than one project, give each window to one project only. The multifaceted entrepreneur article covers why separating contexts matters; the short version is that a window split between two businesses tends to produce two sets of startup costs and half the output. And if your windows are mostly spent creating content, the content creator to founder article is a useful check on whether that content is building a business or just an audience.

Common ways the plan goes wrong

These are the failure points I'm watching for in my own testing. None of them are rare.

  • The brief becomes a conversation. Five minutes of briefing turns into twenty minutes of chatting with the assistant about strategy. Set the brief as read-only: read it, check it, move on.
  • The output is too big. "Launch the offer" isn't a 40-minute output. If you can't picture what done looks like at minute 50, make it smaller.
  • Parking turns into doing. You park a thought, then "quickly" handle it. The if-then plan exists for this moment.
  • The AI joins the block uninvited. A quick question becomes a long thread. If the AI isn't part of the output, its window should be closed.
  • The handoff gets skipped. It's the first thing to go when a window runs over, and the most expensive thing to lose, because the next window then starts from scratch.
  • Feeling productive replaces checking. At the end of the window, the question is simple: did the one output move? If not, the plan needs adjusting, however busy the hour felt.

How to tell if the plan is working

I'm not going to give you a benchmark, because I don't have one I'd stand behind. What I'm watching are a few simple signs you can notice in a couple of weeks without a spreadsheet.

  • Time to first real action. Roughly how long after sitting down do you start the actual output? If the brief is doing its job, this should shrink.
  • Outputs per window. Did the one named output get finished or clearly advanced? A simple yes or no in your log is enough.
  • Re-decided questions. How often do you find yourself debating something you'd already decided? The decision log should make this rarer.
  • Parked items that matter. Look back over the parking list once a week. If important things are landing there repeatedly, they may need their own window.

If those signs aren't moving after a few weeks, change one thing at a time: the size of the output, the length of the make block or how much the AI is involved. As the validation guide says about offers, the evidence that counts is what actually happens, not how it felt.

Woman founder in a warm charcoal knit sweater writing a short handoff note in a notebook at the end of a work session, laptop half closed, evening light on a cream wall
The last five minutes are for the next window: where you stopped, what you decided, and the exact next step.

Frequently asked questions

What is a 60-minute work window plan?

It is a simple structure for one hour of focused work: a few minutes to get briefed on the project, a few minutes to choose one output, about 40 minutes of single-task work, and about 10 minutes to log decisions and prepare the next session. The aim is to spend the hour on the work rather than on deciding what the work is.

How can AI help in a one-hour work session?

AI is most useful at the edges of the hour. It can summarize your own notes into a short brief at the start, hold stray thoughts you park during the work, file your decisions and draft a brief for your next session at the end. It can also help with drafting when a draft is the output. Choosing what to work on and doing the core thinking should stay with you.

Is 60 minutes long enough to do meaningful work?

Yes, if the hour starts with the work. A single, well-defined output such as a first draft of one section, three pricing options or a set of interview questions can usually be finished or clearly advanced in about 40 focused minutes. The planning fallacy research suggests people overestimate how much fits, so choose one output rather than a list.

Should I use AI during the deep work part of the hour?

Only if the output itself is something AI does well, such as a draft you will edit. Research shows large gains on well-defined writing tasks, but a 2025 METR study found experienced developers took about 19 percent longer with AI on complex familiar work while believing they were faster. If AI is not part of the output, keep it closed apart from a parking note.

What is attention residue?

Attention residue is a term from Sophie Leroy's 2009 research. When people switch tasks, part of their attention can stay with the previous task, especially if it was left unfinished, which hurts performance on the new task. Writing down what is still open before you start a work window is a simple way to reduce it.

Does the Pomodoro technique work better than a 40-minute block?

There is limited research comparing specific block lengths. A 2023 study of 87 university students found that fixed break schedules were linked to less fatigue and more concentration than taking breaks at will, with similar work completed in less time. That supports deciding your break point in advance more than any particular number of minutes. Use the length that lets you finish one output without forcing a stop mid-thought.

Is it true it takes 23 minutes to refocus after an interruption?

The 23-minute figure comes from Gloria Mark's observational research on office workers and is widely repeated, often beyond what was measured. Her 2008 experiment found interrupted people worked faster but with more stress and frustration. The safe conclusion is that interruptions have real costs, not that recovery always takes a fixed time.

What should go in an end-of-window handoff note?

Keep it short: the goal of the project, exactly where you stopped, the next concrete step, any decisions you made and why, and any open questions. An AI assistant can turn a rough note into a clean brief for the next session, but check it against your notes before you close.

What if I only have 30 minutes?

Keep the same order and shrink each part: about three minutes for the brief, two to commit to a smaller output, twenty of focused work and five to log and hand off. If time is tight, cut the work block rather than the handoff, because the handoff is what makes the next window start quickly.

Do I need special AI tools for this plan?

No. Any AI assistant that can read notes you give it works for the brief, parking and handoff. What matters more is having your notes, decisions and open loops written down somewhere you control, and a habit of checking the assistant's summaries against those notes.

Where this goes next

This article goes deeper inside the AI second brain theme that started with how to use AI as a second brain and continued with what an AI second brain should do for a founder. It leans on the decision rules in how to decide what to work on first and the weekly rhythm in the founder operating system article. Next in the series: tool vs workflow vs business system, the difference between the three, and why most founders have plenty of the first and not enough of the last.

To get each new article as it's published, join the Mogul Movement. If you'd like decision rules, work-window rhythms and an AI second brain set up around your own business, the Mogul Mastery founding waitlist is open. You can browse every article on the Mogul Maxxing blog, or go back to the homepage for the full series.

Sources

This article is general information about research on attention, planning and AI at work, not professional, medical or business advice. Study findings describe averages in specific settings and may not apply the same way to every person, tool or business. The plan described is being tested in an ongoing build and is not presented as a proven method.