Skip to content
Interface language: English
Imperium / Signal

Communicating change at work: what to tell the team before and after launch

Prepare workplace change communication: reason, tasks, timing, support and current instructions. Check understanding before launch and observe work afterwards.

Article language: English

A baker divides warm bread in a kitchen at night, with navy cupboards and a stone hearth.
In this article

"From Monday we are working differently. Details are in the attachment." The attachment has an attachment of its own, Monday is tomorrow, and the team is trying to find out whether they may still finish work already under way. A change announcement should explain the effect on specific work: what to do, from when, with which materials and where to find help. Announcing a novelty alone does not prepare people to use it.

Start with the agreed scope and the groups affected. Then prepare communication before launch, support for the first tasks and a way to update information after collecting feedback. The message itself matters, but so does consistency between instructions, permissions, examples and decisions. Even the best-written email cannot resolve contradictions between them.

First establish what is actually changing

Imagine a fictional company introducing an AI tool to prepare draft summaries of documents. The change covers selected materials, employee review and a specified place to save the result. It does not allow the tool to make decisions independently. These boundaries should be approved and recorded in the instructions before the team is invited.

Separate an intention from a decision already made. "We want to explore usefulness" means something different from "we are launching a new way to handle every case". If a trial is under way, call it a trial and state its participants and scope. A business experiment card (in Polish) helps identify the assumption to test and the decision the result will support.

Also record what remains a person's responsibility. An employee should know whether they review the content, approve the result, select the source or do all three. The boundaries of a responsible AI agent workflow (in Polish) provide a useful reference. A convenient feature does not itself extend permission to use it in every task.

Group recipients by task, not mailing list

The person preparing a summary needs to know which materials to choose, how to check the result and where to save it. A manager needs to understand work organisation and how to report problems. The support team needs a way to identify an incident. One message may give shared context, but the details should fit these different roles.

For each group, list what it stops doing, what it starts doing, what its members must be able to do before launch and what confirmation it needs. Include people covering for others, working different hours or joining the process later. A message should not depend on attendance at one meeting that cannot later be revisited in a useful form.

Do not assume that no questions means understanding. Someone may not know whether an issue falls within their responsibility or may hesitate to send a concern to too broad a group. Provide a simple channel for questions and name the person who answers. A shared inbox can work well if someone is actually responsible for it.

Explain the reason without inventing miraculous results

The reason for change should relate to a recognised difficulty. In the fictional example, the team may need a more organised first review of materials. That does not yet mean the new method will shorten every task or ease everyone's workload equally. Such effects need to be checked within a defined scope.

Prosci's ADKAR model distinguishes understanding the need for change from knowing how to proceed and being able to apply a new way of working. It is a model of individual change offered by a particular company, not proof that every implementation using the structure succeeds. The distinction is useful: explaining why does not replace preparing someone to do the task.

Do not treat questions as proof of resistance to progress. A comment about missing access or unclear review may reveal a real gap. When describing expected benefits, use the principles of connecting claims with evidence (in Polish). An internal announcement can also promise more than a project can substantiate.

Show an ordinary day after the change

Rather than starting with a feature list, show one recurring task. An employee receives material within the permitted scope, prepares a summary, compares it with the source and saves the result in the right place. At every step, they should know how to recognise correct execution and when to stop.

Add an example outside the scope. Material might need a different handling process or contain information that company rules do not allow in the tool. The instruction should lead to the appropriate alternative, not end with a prohibition. Set the details in your company's own AI use policy. Our guide to preparing that policy (in Polish) can help you organise the rules.

A short comparison of activities before and after the change works well. It need not be a table. You can list what is replaced, what requires a new check and where responsibility moves. If you still cannot describe the difference simply after this exercise, refine the process first.

Connect the message to access and practice

Microsoft Adoption's guide to Copilot champion-led activities connects preparation with a specific audience, a real task, expected behaviour and available support. It is guidance for a particular product. Your own communication can likewise be connected to an action the recipient must perform, rather than assessed solely by event attendance.

Before training, check accounts, roles and the availability of practice material. Someone watching the presenter's screen may understand the demonstration and still be unable to perform the work. Address access in line with an assessment of AI integration permissions (in Polish). Do not grant broad permissions merely to make a demonstration run smoothly.

Prepare an exercise with a correct result and a case needing correction. Choose the level of guidance and independence in training materials (in Polish) so that the demonstration does not replace the participant doing the task. The participant should compare the summary with the source, identify a missing point and choose the next step. Training in verifying AI responses (in Polish) helps distinguish operating a feature from assessing its result. Clicking a button does not automatically provide both skills.

Set the transition point and the treatment of work already started

Give the date, scope and conditions for launch. Explain whether the new method applies to cases started from that point or also to those already under way. If the decision depends on the stage, describe the distinction. An employee should not have to guess whether to finish half a document by the old method and the other half by the new one.

In our example, the company could complete existing work under the previous rules and apply the new trial to a specified category of materials. This is a hypothetical option to check against your own dependencies. What matters is an explicit decision, not a single choice for every organisation.

Aeneia conceptual photography

Add what to do when someone is not ready. What should a person without access, returning from absence or facing a task not covered by the instructions do? Who confirms an exception? A clear mandate for the process owner (in Polish) makes an answer possible without involving the entire hierarchy in every unusual situation.

Prepare answers to real questions

The question list should cover matters that actually affect work: permitted materials, responsibility for the result, time needed, what happens after an error and available support. Do not confine it to questions convenient for the project. "Is this an exciting change?" rarely helps someone trying to find the current instructions.

When an answer is not known, name the person finding it and explain how it will be updated. There is no need to improvise in a meeting. There is a need to return to the question and update the place the team uses. If the answer changes the scope, update your team's instructions and internal project decision log. The guide to keeping a project decision log (in Polish) explains how to structure that record.

Distinguish a technical problem, lack of knowledge and an objection to the decision itself. Each needs a different response. Sending the presentation again will not fix an inactive account, and fixing an account will not explain who approves the result. The report category should route the matter to someone able to clarify it. For reporting errors within a team (in Polish), also check whether people can safely point out a problem and receive a constructive response. An address for feedback alone does not guarantee an important message will reach it.

Keep one current source of information

A message can lead to one implementation page containing scope, instructions, status, questions and contact details. When updating it, show the date and the main difference. Do not force people to compare successive messages to reconstruct which sentence was replaced. Communication should not require a personal archivist at every desk.

Connect that place to the company knowledge architecture (in Polish). Working documents, historical materials and current instructions need clear roles. If one person delivers training from an old PDF, a shared project page alone will not maintain consistency.

For new employees, use the same route in the plan for introducing them to company knowledge (in Polish). They should not have to learn about the change as a story of how something new was once introduced. They need the current way of working and a task that checks their ability to work independently.

Prepare three messages with different jobs

The first message announces the change. For the fictional summarisation project, it can explain the problem being explored, the trial's scope and participants, and how questions will be collected. It should immediately say whether the recipient must already do anything. People outside the trial need to know whether it concerns them now or merely signals a possible later stage. This reduces the chance of people accidentally starting work with a solution still under evaluation.

The second message starts the agreed scope. It includes the date, current instruction, required preparation and route to help. In the example, it might identify the selected kind of materials and the obligation to compare the result with the source. It also separates confirmed availability from matters still awaiting a decision. Launch should not rest on a statement that access has probably already been granted. Check that participants can perform the first task.

The third message sums up the first period and introduces a concrete improvement. It might explain that recurring questions led to clarification of the review instructions. It gives the difference and the group affected. If the trial does not justify expansion, it says so directly, without presenting a postponement as a completed rollout. The message should help further work, not maintain an originally announced mood regardless of new findings.

These messages can lead to one current source of information. Each should nevertheless explain on its own why it has reached the recipient and what is expected. Someone returning from absence then need not reconstruct every event in order. They can see the current state, required task and route to details. History remains available, but does not become compulsory reading before an ordinary task.

Check the message before sending it

Give a draft to someone in the audience and ask a few questions: what will they do differently, from when, with what material, what will they check and where will they report a problem? Do not prompt them while they read. Differences reveal missing information more precisely than asking generally whether the text is clear.

Also check whether the recipient distinguishes a requirement from a suggestion. A voluntary trial and a mandatory process change should be named in accordance with the actual decision. Do not rely on tone alone to convey the distinction. Adjectives such as "modern" and "convenient" do not describe the scope of application.

Prepare a suitable format for people who do not attend the meeting. A short description with an example and instructions may work better than a long recording without signposts. Material should enable the task, not merely confirm that an event took place. Attendance and readiness to work are different outcomes.

After launch, observe work, not just message opens

Collect questions, access errors, difficulties with review and situations where the instructions are insufficient. Check which issues concern the message and which concern the actual process. If the tool does not work, a continuity plan for an AI outage (in Polish) is needed, not a more persuasive paragraph about benefits.

After the first period, conduct a retrospective that leads to a change in practice (in Polish). Choose a specific improvement, assign an owner and check the result. Do not automatically rewrite every communication if clarifying one step is enough. Equally, do not leave a significant gap merely because the message has already been approved.

If an AI rollout needs clearer rules, training and responsibilities, see the scope of work with Aeneia Audit.

Aeneia Audit

Show us the planned task, current instructions and the team's questions. We will discuss the scope of diagnosis and preparation for working with AI, so that the change message rests on actual rules. Prepare the team for a change in the way it works with AI

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