Urgent for five minutes: how interruptions break up a team's day
Follow a request through the reply and the return to the interrupted task. Test which messages can wait and which need someone's attention right away.
Article language: English

In this article
'Do you have five minutes?' Sometimes the answer really does take five minutes. But the person asked was just comparing two offer options, keeping a condition from an attachment in mind and one sentence away from completing a conclusion. Afterwards they must work out where they stopped. The calendar shows a brief bit of help. The work itself also includes returning.
Not every interruption is unnecessary. It can stop an incorrect dispatch, clarify a condition another person needs or prevent work on an outdated scope. The problem starts when a company does not distinguish necessary intervention from the habit of getting every answer immediately. Someone else's urgency then becomes the default order for the whole day.
Two clocks for the same request
The first clock measures a message exchange or conversation. The second concerns the interrupted work: whether and when the person recovered the needed context. There is no single number of minutes to add to every question. Task type, timing and contact content can matter. First observe your team in its actual conditions.
In Mark, Gudith and Klocke's 2008 experiment, time spent on the main task, excluding time spent on the interruptions, was shorter under interrupted conditions than without interruptions; participants reported greater effort and time pressure. The study involved 48 people and a simulated office task. It does not justify the universal claim that interruptions speed up every team's work.
This is a reason not to judge the situation by one task's completion time alone. A company should establish what was displaced, what was not checked and what work needs to be resumed. This review examines the route of questions and their effects on specific tasks.
Measuring work time with AI (in Polish) similarly requires a clear measurement boundary. A minute spent generating an answer does not automatically describe the cost of the whole case. Here we also ask whose work was interrupted, when, and what had to happen afterwards.
Trace one request through to the end
Imagine a team organising workshops for companies. A coordinator is preparing conditions for one engagement. A colleague asks for a filename for another customer. Soon a request arrives to confirm a date, followed by an automatic notification of a new material version. Each message is small. Together they change the order of work.
The first request may reveal difficulty finding information. The second may really be blocking a sales conversation. The third may not be needed until the next material review. If all reach the same person with the same urgency signal, the company makes them reassess priority for every notification.
Choose a few ordinary requests and record their route without rating employees. What triggered the contact? What answer was needed? Did it have to come from this person? Was the lack of an answer stopping someone's work now or only later? This describes the arrangement of tasks. It does not require recording private conversations or tracking every click. If the question concerns the proportion of different activities over a longer period, work sampling (in Polish) can be planned separately; a few traced requests do not provide that distribution.
An internal service catalogue (in Polish) can help route recurring needs. When a team repeatedly asks for the same kind of information, a clear place to find it or a clear role may help. Not every request needs a new procedure. Each should have a sensible route to an answer.
Look at the return as well. Did the coordinator remember which condition she was comparing? Did the message change the original scope? Did she need to ask someone else again? These details explain more than 'conversation: five minutes'. Help may be needed and still benefit from better organisation.
Preserve the request's meaning for the person asking. They may see no other path to an answer or may have received conflicting instructions. A new reporting method that leaves that gap unresolved may simply move contact to another channel. The aim is a workable route for information, not a ranking of people who contact colleagues most often.
In the fictional team, a filename question can be answered by identifying where current material is kept. A date question requires confirmation of a particular availability. One generic 'check the documentation' answer will not resolve both. The responsible person should distinguish difficulty finding data, missing data and a decision no document can make by itself.
Timing information is also a decision
In their 2001 study, Cutrell, Czerwinski and Horvitz examined instant-message notifications during search tasks. Results depended partly on interruption timing. This historical experiment concerned specific computer tasks, not a ready-made settings rule for today's company.
The organisational question is which information must arrive immediately and which can wait for an agreed moment. If an answer deadline matters, give it with the reason. 'I need confirmation before the 14:00 call' tells the recipient more than a red exclamation mark alone, provided the time comes from the real situation rather than a tradition of decorating every message as an alarm.
The fictional workshop team can distinguish an urgent question about a current engagement from an ordinary addition to material for a future meeting. The first needs a recognisable intervention route. For the second, people need to know where to leave the question and when to expect an answer. The rule must be workable on both sides.

It is not enough to declare interruption-free work while simultaneously expecting immediate reactions to everything. That transfers the contradiction to individual employees. The team needs agreement on which work yields in a particular situation and who can make that choice.
When setting a response, distinguish acknowledging a question from answering it. A short acknowledgement can say the right person knows about the case and has identified a next step. It should not pretend to be a solution. If a condition must be checked with someone else, say so. Otherwise the questioner may think they received confirmation while the recipient believes they only promised to return to the subject.
Check whether the same request goes to several people at once. Without a shared reference point, each may begin checking independently and the questioner may receive different answers. This does not prohibit consultation. People need visibility into who leads the clarification and where the confirmed result is recorded, so help does not create several uncoordinated promises.
Some interruptions start without a new message
In a study published in 2011, Dabbish, Mark and Gonzalez also examined self-initiated task switching. It involved observations of 36 people in three work groups from two companies, including two departments of the same firm. Distinguishing self-initiated switching from external triggers helps describe the phenomenon; it is not a diagnosis of a particular person.
Someone may return to a messenger because they await an answer that blocks work. They may also remember several unfinished matters and check each in turn. Before treating this as an individual discipline problem, see whether tasks have a clear next step and a way to get the necessary data.
Too much work started at once (in Polish) adds open contexts. If five matters have unclear states, each can resurface at the next question. Reducing notifications alone will not resolve a dependency the team left without an owner.
Similarly, a company dependent on one expert (in Polish) may keep interrupting that person's work even while everyone says they protect focus. Identify which questions truly need the expert's competence and which arise because information is unavailable to others. Only then choose a change.
The team can help by naming the next check for a waiting matter. In our example the coordinator should know when date confirmation will arrive and what she can do beforehand. She need not reconstruct an unclear dependency every time she opens the messenger. Observe the effect in actual work; writing a new rule alone does not prove the need to return has changed.
Trace the coordinator's return. After the answer about the date, she returns to comparing conditions, but the material was updated in the meantime. She must establish whether she is still using the correct version. Another check may result from a real change rather than the interruption itself. Preserve that reason in the description rather than attributing the whole difference to one notification.
If the answer was incomplete, the original contact may return a few minutes later. Three short questions may then be one unfinished matter. Find what prevented a useful first answer: missing data, a different question scope or the absence of the right expertise. The number of messages alone cannot determine which occurred.
Also recognise contact that actually saved later work. Someone noticed the wrong scope and alerted the person preparing an offer. The interruption may have been needed at that moment. Any change should preserve such a signal. Otherwise the team may improve the notification-count metric while doing more corrections after an outdated document is finished.
These cases help select the aim. The team may want fewer repeated questions about readily available data and also maintain quick contact about a confirmed discrepancy. These compatible needs require a clear rule for asker and recipient. A good result need not mean silence. It should mean the right information reaches the right task at a time that fits its need.
Test fewer interruptions in one type of case
Instead of redesigning the company's whole day, choose a recurring contact type. In our example it could be questions about the current material version. Establish a clear information location and a way to report discrepancies. Keep an urgent intervention route where an error could reach the customer. Then check whether the need for direct questions actually changed.
Do not measure only messages sent. Fewer contacts may mean better access to information, but could also mean cases wait unanswered. Check the result: did the person find the correct material, did work move forward and was a new ambiguity created? Silence in the messenger is not by itself a measure of work well done.
Some questions may result from an unhelpful case status. If an update says only 'in progress', the recipient still cannot tell whether to wait, supply data or help remove a blocker. Improving the status content may be more appropriate than another ban on messaging the team. For ways to make a status useful without another meeting, see the guide to team status updates (in Polish).
Cooperation across time zones needs clear expectations too. Availability between Poland and Asia should follow the agreed working method, not fixed habits attributed to nationalities. A recipient needs to know when and how to raise a matter so the other side can respond appropriately. The guide to cooperation across Poland and Asia (in Polish) discusses how to agree on communication across time zones.
If AI generates some notifications, trace one from source to action. Does it report an actual event or only a new draft? Must the recipient react, and if so, what should they check? A selected signal might go to a quiet review and another to urgent assessment. That is a decision for a specific scope, requiring confirmation of conditions and a check of the outcome.
You can discuss your needs concerning AI in such a flow and a possible scope of cooperation with Aeneia Audit. For the conversation, prepare an example: where a notification originates, who checks the answer and what action follows it.









