Skip to content
Interface language: English
Imperium / Signal

Before changing an expert's work, see an ordinary day

When do instructions and an employee's account need direct observation? Follow one case and recognise the limits of the conclusions.

Article language: English

A woman holds a sheet of paper beside an ornate red cabinet of drawers.
In this article

The instructions contain three steps: receive the material, pass it to the person doing the work, record the deadline. At the desk, the employee opens an earlier video version between the first and second steps, checks the client's note and calls with one question. Only then does she hand over the task. That part, invisible in the instructions, may determine whether the next person receives the right material.

Imagine a company preparing captions for training videos. The manager wants to improve order intake and believes coordination mainly means moving data from messages into a system. An experienced coordinator says most of her time goes on “clarifying things”. Both may honestly describe what they see. Before changing responsibilities or introducing automation, the company should establish what lies behind that general phrase.

The instructions, the employee's account and watching one case with the employee aware of the observation provide different information. None needs to be treated as the only true perspective. The company needs to know which question each helps answer and what it does not yet prove. Observing ordinary work can reveal an important unknown. It does not by itself justify assessing the productivity of the whole company or everyone doing similar tasks.

What disappears between the names of steps

A formal description is a shorthand. “Pass on the material” may mean simply sending a file, or it may require checking whether the file is the current version. In our example the client sends a video whose name contains “final”. In the same message they say the final shot will still change. The coordinator need not know immediately what to do. She must notice the discrepancy and obtain an answer that allows the task to be handed over.

If the manager counts only applications opened and messages sent, they may see unnecessary switching between tools. If they ask why the check matters, they learn that the earlier version explains the client's reference to the final part. The same activity may look like wasted time or a necessary attempt to understand the material. The assessment depends on knowing its role in the task.

In their 1991 article, Brown and Duguid point to the difference between formal descriptions of work and practice. Among other things, they analyse an earlier ethnographic study of service technicians. This supports understanding specific actions before simplifying them; it does not prove every undocumented activity is necessary and should remain unchanged.

The fictional company should separate two questions. First, is the coordinator doing work the manager did not account for? Second, is the current way of doing it the best one? A yes to the first does not answer the second. Perhaps the necessary clarification can be obtained earlier or passed on differently. First, however, the company must identify what clarification is needed at all.

A service map covering visible work and the back office (in Polish) can help at such a point. Do not draw more boxes merely to preserve everything observed. For the decision to change, what matters is which activity affects the material, responsibility or the next choice. The map should help ask about that effect.

One ordinary case can change the manager's question

The coordinator might describe the order as “I checked the version because the client mixed up the files”. Watching the process, the manager may see something more complex: the client did not mix up the attachment but understood readiness differently. The video was ready for consultation, not for work on final captions. That changes the proposed solution. Preventing an incorrect attachment from being sent would not remove the difference in meaning.

Here observation does more than list additional activities. It helps check whether the company named the problem correctly. Instead of “the coordinator takes too long to pass files”, the question becomes: “Who recognises that material is ready for the next kind of work, and on what basis?” That can concern the client, the person doing the next task and how scope is agreed, not just one person's speed.

Do not guess the meaning of every gesture. The employee may open an old message for a reason different from the observer's assumption. She may also choose a route that was exceptional that day. If an important decision depends on that moment, clarify it with the person doing the work. “She returned to the previous message” and “she did not know the right version” are different kinds of information.

The GOV.UK guide to contextual research and observation describes learning about activity in its everyday setting with the materials and tools normally used. It also notes that questions and explanations of successive steps change the natural course. This is an important limit: a situation in which an employee explains every action to an observer is not an ordinary day reproduced without outside influence.

The manager is not therefore obliged to design a large study personally. They do need to describe honestly what they learned. If they watched one jointly selected case and asked about several moments, that is how the basis for a decision should be presented. “We checked how coordination works” may suggest a much broader scope than was actually covered.

A case reveals a possibility, not its frequency

One difficult video is enough to notice that checking material readiness sometimes requires human judgement. It does not yet say what proportion of orders have that character. If the company changes its entire service model, it needs to know whether it saw an exception, a recurring type of case or a daily duty. Each has a different effect on the sense of the planned change.

Aeneia conceptual photography

Work sampling to show the structure of time (in Polish) may help when the question concerns the share of activity types over a longer period. That is a separate task from examining reasoning in one case. Do not turn a detailed story into a percentage of time merely because it was well documented and memorable.

Likewise, observing one person does not prove only that person can do the work. She may use her own shortcuts, while another employee solves a similar case differently. If the company is concerned about dependence on one person (in Polish), it needs to test whether the task can be taken over and under what cover arrangements. Seeing a capable expert is not such a test.

In our fictional company the manager may decide, before automation, to learn about an ordinary complete order and a case with a late change to the material. They want to see whether the same readiness condition appears in different situations. Further observation should fill a specific gap on which the choice of solution depends.

The initial hypothesis may also prove correct: some data transfer really may not require additional judgement. Looking at the work then helps define the boundary of that task. The company can simplify that part while preserving other actions needed to clarify the material. Instead of choosing between full automation and leaving everything alone, it gets a narrower, justified scope of change.

Connect the scope of further study to the decision the company must make now. If it only wants to copy confirmed data into a system, it must establish how confirmation is recognised. If it wants to remove the coordinator's readiness judgement, it needs to understand the grounds for that judgement much better. The two intentions do not require an identical picture of the whole role. In the second, the omitted unknown concerns precisely the function the company intends to change.

The manager may therefore narrow the first step: leave judgement with the coordinator and separately test transfer of already approved information. That choice should be explicit. Do not later present the trial as proof that the tool can also take over assessment of the material. A different part of the work was tested. A narrow start is useful precisely when the broader question it has not answered is known.

If the next decision depends on the unexplained part, the company should allocate time to understand it or limit its change plan. Apparent acceleration achieved by omitting that cost can leave the next worker with a task for which they have neither information nor authority. In our example someone would still need to agree whether the final shot can be processed. Renaming the role would not supply the answer.

Not every workaround is a good solution

An expert may perform a useful check by an inconvenient route. Suppose the coordinator identifies the video version by comparing names in several messages because the current material has no unambiguous label. That effort currently protects later work. The company can improve labelling instead of preserving an obligation to compare every message manually. It retains the necessary information, not necessarily the existing effort.

Another workaround may pass the problem further along. The coordinator adds her own assumption to avoid delaying the handover, and the next worker must return to the client with a question. If observation ends when the file is sent, the manager sees quick closure. Only checking what the next person received shows the cost of the shortcut. The boundary of observed work affects the assessment of the solution.

Also ask about evidence for the problem's causes (in Polish). “Clients always send unclear materials” may be a convenient explanation, but it does not explain why this order was accepted in that form. The company's own form, an ambiguous promise or a different kind of agreement may matter. Consider them instead of stopping at the first convincing description.

The employee should know that their work is being observed and understand its purpose. If the company asks her to show difficulties and then treats every uncertainty revealed as grounds for an individual ranking, it changes the meaning of the situation. Here the task is to assess a planned change in work. What may be shown, recorded and passed on should be agreed appropriately to the actual data and company conditions.

A hypothetical AI use case

Suppose AI prepares a draft video transcript and the coordinator passes it on. The formal description may say “generate and check”. In practice the company needs to know how she identifies the right input, which places require returning to the recording and what her confirmation means. Without that, an automation project may remove an activity whose role nobody had named.

The case can be compared with a register of data flows used by AI (in Polish). Actual work may reveal material passing between tools that the formal record does not include. Only then does the company have a basis to assess the specific input, output and checking rules. There is no need to begin with a declaration about transforming the whole organisation.

Aeneia Audit can help trace such AI use with the people doing the work: organise sources, how the result is used and responsibility for checking. A good starting point is a task whose actual course the team can show and whose unknown it can name. That allows the tool to be assessed in the work context for which the company wants to use it.

After such discovery the manager's decision may be more modest than the original plan, but better defined. They know which part can be simplified, which check needs to remain and what still needs to be understood. The company can change its method while preserving necessary checks. Its ordinary day supplies the knowledge needed for change to address a real task rather than only three convenient names in an instruction.

Aeneia / Aeneia editorial

Editorial article. Source links are included in the text.

About Aeneia

Brands we have worked with

  • L’Oréal
  • Garnier
  • Netflix
  • Carrefour
  • Warner Bros. Discovery
  • Biedronka
  • HBO Max
  • Wolt
  • Żabka
  • COSRX
  • Embassy of China in Poland
  • Mixa
Aeneia / Initium

From knowledge to a decision

Tell us what you want to put in order. We will agree on where to start.

Contact Aeneia

We partner with and use solutions from:

  • Google
  • OpenAI
  • Microsoft
  • Stripe
  • Anthropic
  • Linear