Hybrid meetings: half the team cannot be an accessory to the speakerphone
Give remote participants access to the discussion, shared material and decisions in a hybrid meeting, and check AI summaries honestly.
Article language: English

In this article
“Can everyone hear us?” A “yes” comes from the speaker, so the meeting begins. Soon someone walks to the whiteboard. Two people at the table exchange a quiet comment, a third points to a piece of paper, and a remote participant types a question in chat. Nobody notices it. The connection works. Shared work may not.
In a hybrid meeting some people are together and others join remotely. This creates different routes into participation, but does not make them equally influential. A sentence at the table immediately changes the conversation. A chat question may wait until the thread ends. Before it is read, the group may choose the very solution it concerned.
Imagine a small museum network agreeing how to handle group visits. Four people meet at headquarters and two employees from local sites join online. Everyone knows the goal: decide how to pass on a request to change a visit time. Everyone speaks Polish. Even so, the people in direct contact with groups may find it hardest to contribute their experience. The problem does not begin with the agenda or vocabulary. It begins with what each side can see, do and challenge.
Seeing the whiteboard is not yet access to the task
At the fictional headquarters someone draws two ways to pass on a request. People in the room see the arrows and hear brief additions. On the remote screen the board is small and its author obscures part of it. After “which option do we choose?”, a remote participant is silent. We do not know whether she supports the proposal, cannot see the difference or is trying to read an annotation. Inferring a conclusion from silence alone would be guessing.
Sharing a document can help if work actually happens there. It will not help if the shared file is only a clean copy while all decisions still emerge at the physical board. A participant needs access to options while they can still be changed. Access to a finished summary provides knowledge of the result, not participation in making it.
Start with the actions people must perform. Must they view one item, add a case from their site, compare two service paths or raise a reservation before a choice? Check separately how a person in the room and a person elsewhere will perform each action. The same setup need not support a presentation and parallel idea generation equally well.
Research by Saatçi and colleagues from 2019, involving two global software companies in Europe, supports this distinction. Observations, interviews and recording analysis showed varied meetings and unequal interaction between rooms and remote participants. It describes particular workplaces, not a current equipment test or proof that one application is superior.
In the museum example, put both options in material everyone can access and refer to them by name. If someone adds an exception at the board, it should also enter the shared view of the case. People on the other side should know what changed the proposal and what they can respond to.
Who notices someone waiting to speak?
Conversation around one table uses signals easily missed over a connection: a glance, preparation to speak, a pointing hand. A remote person may see several small faces or one room panorama. If they start speaking, they interrupt someone. If they wait, the thread closes. “Feel free to interrupt” places the cost of repairing the meeting arrangement on them.
Before the conversation agree how to signal a question and who watches for it. If chat is the place for questions, somebody must actually read it while work is happening. If people signal by voice, the facilitator must know when to pause the exchange at the table. More roles are not inherently needed. The chosen participation method needs a recipient and a way to affect the discussion.
Calling someone by name is not always an invitation to meaningful participation. “Anna, do you agree?” after the discussion asks for judgment on a finished result. “Anna, which case from your site does not fit these options?” lets her contribute knowledge to the task. In our example the second might reveal that a group organiser requests a change after some visitors are already travelling.
At that moment “we'll add it later” is not enough. If the case undermines the chosen service path, it must return to the comparison. Otherwise the company records the practitioner's attendance without using the knowledge for which it invited her. The facilitator's job is to keep contributions connected to the question, not merely divide equal portions of time.
For creative work, people can prepare individual ideas and later develop them together (in Polish). A quick exchange in the room need not be the only entry into discussion. Remote participants still need access to shared material, but do not have to compete for speaking order while recording their first proposals.
Do not measure participation only by speaking time
One person may speak for a long time to present material. Another contributes one sentence revealing an important condition. A third reconstructs cases from the past month and prepares a comparison before the meeting. Minutes at a microphone do not automatically measure the value of these contributions. Equal speaking time would be convenient for a chart but not necessarily for the museums' task.

A 2024 study by Hosseinkashi and colleagues compared survey ratings of effectiveness and inclusion with available meeting features. The authors describe relationships between measures and difficulties measuring subjective experience. Such correlation does not show that changing one feature will cause the desired result in any team.
For the fictional network, work-related questions are better than a speaking-time contest. Could the site employee show an exception before the choice? Were the authors of the comparison visible in discussion? Was a reservation recorded with the response? After a connection interruption, was it clear which part the participant missed? These questions expose concrete barriers.
A difficulty may overlap with unequal ease in the working language (in Polish). Fast side conversations can then be especially difficult to reconstruct. But do not explain every lack of participation as language proficiency. In the museum example everyone uses the same language and the problem remains. First check actual access to conversation and material.
It also helps to recognise contributions not visible during presentation (in Polish). If the facilitator presents a comparison prepared by site employees, name its origin. If the summary attributes the whole analysis to the person who read it out, the meeting preserves a false picture of the work. This is especially easy when the authors appear in small windows on a screen.
A guest can enter the call but not the file
Suppose the museums invite an external group coordinator to the next discussion. She joins the call, but the shared document is available only to network employees. For the organiser everything looks fine: screen sharing has no error. The guest can see the material but cannot add her case or inspect an earlier note. This differs from what the internal team rehearsed.
Before work begins, test access from the guest's position. Is she using a phone or computer? Can she open material independently of screen sharing? Does she need to work in it or only respond to what is displayed? The answers permit an accessible way to contribute. They do not require giving everyone identical rights to every company document.
In our example the coordinator could give her case to the person recording shared findings. That person must read the record back and confirm it captures her point. “We've got it” leaves the guest unable to check whether her words were understood. For a long point or one covering several options, prepare accessible material earlier instead of improvising as the meeting ends.
Separate difficulty operating a tool from absence of an idea. Someone unable to place a note on the board in time may have had an answer ready. Moving on omits the proposal even though everyone formally received the same task. After such a trial fix the way material can be contributed before judging who produced the most ideas.
When the connection stops being enough
Return to the moment of choosing an option. One participant loses audio. People in the room can continue, but must decide whether the planned task can be done without her knowledge. If she only needed to hear information, the material could be shared later. If she needed to check a service condition at her site, the group lacks that answer. Continuing the conversation does not create the missing check.
The group can defer the decision while using remaining time to compare matters independent of the missing contribution. It can also return to the question in writing if the task permits. The result must not say “everyone agreed” when some people could not respond to the last change. An honest record of no agreement or no answer gives the team a further task.
This links the hybrid arrangement to preparing a decision meeting (in Polish). Authority to choose and conditions needed for assessment should not vanish with a microphone failure. Even if one person decides, they should know which information is missing and why it matters. “She was on the call” does not prove she could perform her part.
Keep the case as guidance for the next meeting. Perhaps the main comparison needs to be sent beforehand. Perhaps vague references to “here”, rather than image quality, are the issue. Perhaps one person cannot simultaneously lead a difficult discussion and read questions. Identifying the actual constraint lets the team choose a change suited to the task.
An AI summary cannot recover an unheard question
A tool can prepare a polished note from what the recording or meeting record captured. It cannot restore a comment that was never heard or the meaning of an arrow visible only at the table. In the hypothetical museum network, “a shared route was agreed” can look credible even though a remote participant was trying to raise a case that undermined it. Fluent text does not repair a missing conversation.
An audit of AI used for meeting records (in Polish) should therefore consider available input and how the result is checked. For a specific use, compare the summary with findings, identify unresolved questions and decide who confirms the record. Aeneia Audit covers assessment of AI use, information flow and responsibility for the result.
The museums' next meeting can be prepared around one verifiable task: each site responds to the same two options, and its reservations enter the comparison before a choice. The question of hybrid setup quality then has a concrete answer. Hearing headquarters is not enough. People must be able to change what they are working on together.








