A fundamental principle of any system designed to improve productivity is to finish things properly. If we consider productivity to be a matter of managing energy, every task left hanging represents a little attention that will have to be mobilized later to work out where we left off.

Ideally, then, we would finish one task before moving on to the next. Reality is rarely so cooperative. A call comes in, a meeting starts, an emergency crops up, or the available time simply runs out sooner than expected. We are not always in control of our schedules, and some tasks take several days of work.

Rather than making complete closure an impossible rule to follow, I prefer to talk about switching tasks well. This is not just about stopping what you are doing. It is about leaving the subject in a state clear enough that the interruption no longer occupies your mind and returning to it is easy.

The cost of leaving a task open

An unfinished task is not necessarily a problem. Writing an article, preparing a file, or renovating a room naturally takes several sessions. What costs energy is leaving the subject without knowing exactly what has been done, what remains to be done, and where to begin again.

This creates a form of mental noise. During the next task, part of your mind continues to wonder whether you should have replied to that message, saved that document, or kept open the tab containing useful information. Even when you are not consciously thinking about it, you hesitate to close your tools for fear of losing the context.

The same thing happens in physical spaces. Starting to empty the dishwasher, leaving half the dishes on the counter, and then moving on does not just create an unfinished task. It also clutters the kitchen and makes the next action less obvious. Later, you will have to work out what is clean, what has been put away, and what remains to be done.

The dozens-of-open-tabs syndrome is the most visible digital version of this. Every tab seems to need to stay there “for later” because it represents an intention that has neither been completed nor properly recorded. The computer ends up displaying the state of all our indecision.

Interrupt well instead of finishing everything

It is tempting to address this problem with a strict rule: never start a new task before finishing the previous one. This works for some small actions but quickly becomes unrealistic for everything else. A task may depend on someone else, take longer than expected, or have to make way for a genuine scheduling constraint.

The alternative is to give the session a real ending, even when the task itself continues. Before switching subjects, I try to answer four simple questions:

  • What has actually just been accomplished?
  • What state is the task in now?
  • What is the next concrete action?
  • What needs to be kept so I can resume without searching?

The answers can fit into one or two sentences. The aim is not to write a detailed report every time you are interrupted, but to create a reliable place to resume. “First section reviewed; resume at the paragraph about interruptions and check the dishwasher example” is already enough to avoid several minutes of rereading and hesitation.

This closure may also lead to a change in status. If I am now waiting for a response, the task moves to “Waiting.” If I decide not to return to it before a specific date, it moves to “Snoozed.” If it remains actionable, its next action simply needs to be up to date.

Mark a milestone in the Kanban board

This principle is directly rooted in how I manage my Kanban board. After making meaningful progress, I update the date of the last activity even if the task is neither finished nor moved to another column. This brings it back among recently active subjects, and its history shows that a real milestone has been reached.

You have to remain honest with yourself. Opening a card, rereading its title, and changing its date does not count as progress. This feature should not become a way of artificially making it look as if everything is moving forward. I update the date when the task is objectively in a different state: a section has been produced, a decision has been made, an avenue has been ruled out, or the next action has become more precise.

This record has two useful effects. First, it makes the work already done tangible, even when no task has yet met its definition of “done.” Second, it makes it easier to resume because the card explains where the subject stopped instead of presenting an instruction that is now obsolete.

A long task does not need to be treated as a single, indivisible block. It can have several intermediate endings. Each one properly closes a session and prepares the next.

Close and put things away to free your attention

When I switch tasks, I also try to close or put away as much as possible. I save the document, turn a useful page into a link in the task, close tabs I no longer need, and put objects back where they belong. This can take a few minutes, but it gives the interruption a clear structure.

It can feel like a waste of time. After all, the page is already open, the cleaning supplies are out, and the equipment may be used again later. Leaving everything in place can therefore seem more efficient.

But this calculation overlooks the cost of resuming. An environment left hanging rarely preserves all the context you will need. A few hours or days later, you no longer know why three similar pages were open or what the object in the middle of the room was meant for. You then have to reconstruct the intention from incomplete clues.

Closing does not mean deleting. On the contrary, you first need to move useful information to a reliable place: the task description, a note, a bookmark, or the project folder. Once that is done, the temporary tools can disappear without concern. The context is no longer stored in a cluttered workspace or the fragile memory of a browser session.

Taking inspiration from commits and branches

Software development provides a good illustration of this discipline. A developer avoids leaving code in an incomprehensible state for long. When a coherent set of changes is ready, they create a commit with a message that clearly describes what was done.

The complete project is not finished, however. The commit simply marks a clean, identifiable state. It becomes possible to resume the work, understand its history, or revisit a decision without having to interpret a mass of indistinct changes.

The concept of a branch is also interesting. Creating a branch means defining an intention: fixing a problem, testing an idea, or developing a feature. The changes belong to that specific piece of work instead of immediately becoming mixed in with everything else.

We can apply the same rigor to everyday life. Before interrupting a piece of work, I can ask what its “commit message” would be: what coherent unit have I completed, and what would someone need to know to pick up from here? I can also explicitly name the intention of the next session, as if I were opening a branch dedicated to a specific objective.

The analogy has its limits, of course. This is not about documenting how you put away the dishes as if it were a software project. It simply reminds us that work can remain unfinished while still being left in a clean, understandable, and reversible state.

The forms of multitasking I allow myself

Avoiding multitasking does not mean you have to devote your full attention to every action of the day. Some activities are repetitive and routine enough to leave part of the mind available. When I iron, tidy up, or perform a mechanical task, I can listen to a podcast or leave a video playing in the background.

I do not really consider this to be two demanding tasks performed in parallel. One activity requires few decisions, while the other occupies the remaining mental space. The combination stops being useful as soon as I have to reread the same passage several times, rewind the podcast, or make mistakes in the main task.

The right criterion is therefore not an absolute ban, but the level of attention required. Writing a delicate message while following a video generally produces two mediocre experiences. Folding laundry while listening to a conversation, on the other hand, can work without difficulty.

Use maintenance tasks as a breather

I also use cleaning and maintenance tasks as a thread running through the day when I work from home: vacuuming, washing dishes, dusting, doing laundry, or ironing. I work on a demanding task, then deliberately interrupt it at a clean milestone to spend a few minutes returning to this household routine.

This alternation lets me continue thinking about subjects that require structure or brainstorming without remaining motionless in front of a screen. The physical activity is simple and concrete and often has an obvious endpoint. It provides a breather while still moving something useful forward.

The distinction matters: I do not abandon a complex task in the middle of a sentence whenever I see a plate that needs putting away. I choose the moment of transition, note where to resume, and then genuinely switch modes. Maintenance becomes a deliberate break, not a new source of constant interruptions.

This technique will not suit everyone or every task. If the interruption breaks a good state of focus, it is often better to preserve it. But when your thinking goes in circles or your energy dips, a routine action can help you step back without opening another intellectual project.

Finish the session to make resuming easier

The best scenario is still to finish one task before starting another. A completed task frees up space, reduces the number of active subjects, and avoids any cost of resuming. But making this scenario a permanent requirement would mainly mean ignoring the reality of long projects and unpredictable days.

When you do have to stop, the point is not to let an accidental interruption decide the state in which the work is left. Marking the milestone you reached, updating the next action, recording useful information, and tidying your environment is often enough to create genuine closure.

You do not always finish the task. But you can finish the session properly. The difference may seem small, but it makes context switching much less costly and lets you quickly pick up the thread when it is time to resume.