Before an AI Detour, Leave a Note for the Work You Pause

A creator pauses a newsletter draft to ask AI about a possible headline. That question leads to an idea for a workshop, then a new product description. Returning to the newsletter, they must reconstruct which paragraph was unfinished and which claim still needed checking. This hypothetical detour may have produced useful material, while leaving the original work harder to pick up.

Before changing tasks, a short note can make the intended return explicit. Research on attention and interruptions offers a reason to consider that habit. Applying it to AI-assisted work requires care: the studies discussed here examined human task transitions, rather than today’s chatbot workflows.

What attention residue describes

In her 2009 paper on task switching, Sophie Leroy reports two experiments showing that thoughts about a previous task can persist after people move to another. Leaving work unfinished made disengagement difficult and subsequent performance suffered. Completion and time pressure were part of the experimental design; simply finishing the first task did not, by itself, ensure an effective transition.

Attention residue describes that continuing involvement with earlier work. It gives a more precise question than whether a person has too many tabs open: what are they still thinking about while trying to do the next task? A closed window and a completed mental transition are different things.

A return plan can matter while you are away

Leroy and Theresa Glomb’s Tasks Interrupted, published online on 11 May 2018, focused on performance during the interrupting task. Across four studies, anticipating time pressure when returning to unfinished work was associated with attention residue and poorer performance on the intervening demand. Briefly reflecting on and planning the return mitigated that effect in the studied conditions.

The result concerns attention available for the task entered during the interruption. It should not be rewritten as a universal recovery-time estimate. Nor does it establish that an AI-generated note, a chosen chatbot detour and an imposed workplace interruption have identical effects.

My practical inference is to leave a clear return point before a deliberate switch to AI. The routine below is an editorial adaptation, not a validated AI intervention or a guaranteed increase in output.

Leave a note for the particular work you pause

Use a note beside the actual task, where it will be easy to find when you reopen the work. Keep it short enough to maintain, but specific enough to guide the first action back. Four fields provide a starting point:

• Location: the document, section or item you were working on.
• State: what is already decided and what remains unresolved.
• Next action: one concrete step to continue the work.
• Return point: when or after which bounded activity you intend to resume.

For the fictional newsletter, the note might read: “Draft, paragraph three. Keep the opening example. The workshop date still needs checking against the organiser’s confirmation. Next, verify that date and finish the paragraph. Return after comparing three headline options.”

That example records an unfinished decision without pretending it has been resolved. It also makes the proposed detour finite. A note saying only “finish newsletter” would leave the creator with most of the reconstruction still to do.

If you ask AI to draft the note, compare it with the actual document and your decisions before relying on it. Separate an assistant’s suggested next step from the step you have chosen. The purpose is to preserve a usable working state.

Give the AI detour a job and an ending

For this creator, the immediate question is which headline fits the existing newsletter. A request for three alternatives and a brief explanation of their differences keeps that purpose visible. A request to develop an entire promotional campaign would create a different piece of work.

Neither choice is inherently wrong. The useful decision is whether to make that switch now. If a worthwhile workshop idea appears, record where it belongs and decide whether it warrants replacing the current priority. Keep that choice separate from the mere arrival of another interesting suggestion.

At the intended stopping point, review what the detour produced. Save the selected headline or the unresolved question with the newsletter. Place any separate project idea in a location where you can assess it later, rather than folding it into the newsletter’s completion criteria.

Resume from the work and the note together

On returning, open the document and inspect the recorded location. Check whether anything has changed, then perform or revise the next action. The note is a prompt for reorientation, rather than an instruction to ignore new information.

If another person edited the paragraph while you were away, update the note to reflect that state. If the date remains unverified, retain that uncertainty. Resuming work should not turn a pending check into an assumed fact merely because it appeared in an earlier plan.

This is also a useful boundary between a conversation history and a working document. Decide which record contains the current draft, which contains suggestions, and which decisions you have actually accepted. That is an organisational choice to make explicit for the project.

Assess the whole experience, not only speed

Gloria Mark, Daniela Gudith and Ulrich Klocke’s CHI 2008 interruption study found that participants completed interrupted work faster, with no observed difference in quality, but reported greater stress, frustration, time pressure and effort. The experiment illustrates why speed alone can miss part of an interruption’s cost. It does not measure the effects of current AI tools.

If you try the proposed note-taking routine, consider whether it helps you identify the next step, preserves unresolved checks and keeps additional ideas from quietly expanding the task. Also notice whether maintaining the note itself becomes burdensome. Adapt it to the work rather than judging success by how many fields you fill.

A few personal observations can help you decide whether to keep the routine. They cannot establish a general causal effect: task difficulty, deadlines and the reasons for switching may all differ between occasions.

Make room for necessary interruptions

A customer question or a colleague’s blocker may deserve attention before the current task is finished. Some switches are chosen; others arrive with little warning. A return note is one possible support within that reality, rather than an expectation that every interruption can be neatly scheduled.

For teams, the wider design matters too. MaryChuks.com’s guide to response expectations and digital presenteeism explores how to distinguish genuinely urgent requests from routine messages. An individual note should sit alongside clear priorities and workable handovers.

The useful question before an AI detour is concrete: what will I need to know when I return to this piece of work? Record that answer, give the detour a purpose and revisit the actual task afterwards. The note preserves a decision about where your attention is meant to go next.

Sources

Primary research checked on 27 October 2026. The creator and newsletter are hypothetical; the AI-specific routine is an editorial proposal.

• Leroy: task transitions and attention residue — 2009, Organizational Behavior and Human Decision Processes.
• Leroy and Glomb: Tasks Interrupted — online publication 11 May 2018, Organization Science.
• Mark, Gudith and Klocke: interrupted work, speed and stress — CHI 2008, author-hosted paper.