AI is beginning to move from the edge of conversation into the conversation itself. It can listen, summarise, retrieve information, reflect patterns, and sometimes speak into the flow. This creates new possibilities for collective sense-making, but only if its role is clear and the human conversation remains central.
One interesting question now arising is this: what happens when AI does not simply prepare us for conversations or analyze them afterward, but actually joins the conversation itself?
Until recently, that sounded slightly futuristic. Now it is becoming technically feasible.
Credit: I would like to thank my good friend Bruce Lloyd
for creating1 this NotebookLM video.
In the context of Conversational Leadership, this raises intriguing possibilities but also important questions. AI can become a resource for a group, helping people remember, reflect, connect ideas, and notice patterns. But it can also intrude, distract, dominate, or quietly shift authority away from the human participants.
The key question is not simply whether AI can join the conversation. It is what role it should play, who decides that role, and how we ensure that the quality of the human conversation is strengthened rather than weakened.
AI as a silent listener
The simplest form is an AI system that joins a Zoom or Teams meeting quietly in the background.
It listens to the conversation, transcribes it, and may also identify themes, key points, and unresolved questions during the meeting. Participants might glance at a shared screen that shows emerging summaries, points of agreement, questions raised, or issues that have not yet been addressed.
In this role, AI does not interrupt the conversation. It helps the group see something of what is happening as the conversation unfolds.
This can support better sense-making, especially in complex discussions where many ideas are circulating at once. It can help the group pause and ask: What are we noticing? What are we circling around? What has not yet been said?
But even a silent AI is not neutral. If it is listening, people need to know that it is listening. They need to know whether it is recording, transcribing, summarising, storing, or sharing anything beyond the meeting. Otherwise, the presence of AI may change what people feel able to say.
AI as a conversational aide
A slightly more active role is one in which AI acts as a conversational assistant.
Participants might ask questions during the meeting, such as:
- Can you summarise the key points so far?
- What options have we discussed?
- Where do we seem to agree?
- What issues have we not addressed yet?
- What questions are still open?
In this role, AI becomes a shared resource for the group. It helps participants step back and see the conversation more clearly.
It functions a little like a collective memory. This could be especially useful in longer or more complex conversations, where people may lose track of what has been said, what has been agreed, and what remains unresolved.
But again, care is needed. A summary is not the same as understanding. AI may produce a neat account of the discussion, but that account may miss hesitation, discomfort, minority views, emotional undercurrents, or things that were implied but not spoken.
So any AI summary should be treated as a draft for human correction, not as the official truth of what happened.
AI as a reflective mirror
AI might also help groups reflect on the dynamics of the conversation itself.
For example, it might quietly flag patterns such as:
- certain voices dominating the discussion
- topics that keep resurfacing
- questions that are repeatedly avoided
- points where the group seems confused or divided
- ideas that have been mentioned but not explored
Used carefully, this kind of feedback could help groups become more aware of how they are thinking together.
In other words, AI becomes a mirror for the conversation.
This could be valuable, but it is also delicate. Feedback about who is speaking, who is silent, what is being avoided, or where the group is stuck touches on trust and power. If handled badly, it could feel like surveillance rather than support.
This is why the role of the host or facilitator matters. Someone needs to decide when such feedback is useful, how to introduce it, and whether the group is ready to hear it.
AI as a knowledge resource
Another possibility is that AI can be consulted in real time for background knowledge.
In the middle of a discussion, someone might ask:
- What do we know about this from previous projects?
- What does the research say about this approach?
- What did we decide about something similar last year?
- Are there examples from elsewhere that might help us think about this?
AI could quickly retrieve relevant information from internal documents, previous meetings, or external sources.
Instead of stopping the meeting to research something, the group can continue the conversation while drawing on additional material.
This could be powerful. But it also carries a familiar risk: the conversation can quietly become centred on what the AI knows, rather than what the people in the room are trying to understand.
The question “What do we think?” can become “What does the AI say?”
That is a subtle but important change. AI can bring useful information into a conversation, but it should not replace the human work of interpretation, judgement, and decision.
AI in face-to-face conversations
These possibilities are not limited to online meetings.
In a physical meeting room, an AI assistant could listen through a device on the table and provide summaries or responses through a shared screen, a laptop, or a participant’s phone.
Even in a one-to-one conversation, someone might briefly consult an AI assistant. In a corridor conversation, it might be as simple as asking:
“Quick check, what were the three options we discussed in yesterday’s meeting?”
The AI serves as a conversational companion that can be consulted when needed.
But face-to-face settings make the issue of consent even more important. In an online meeting, an AI presence may be visible on the screen. In a physical setting, it may be less obvious. People should not have to guess whether an AI is listening.
Before AI joins a conversation, the group needs some simple shared understanding. What is the AI doing? Is it listening only? Is it recording? Is it transcribing? Is anything being stored? Who can access it? Can anyone ask for it to be paused?
These are not just technical or legal questions. They shape the trust in the room.
The risk of conversational intrusion
There is a real risk that AI could intrude into the human flow of conversation.
If it interrupts too often, provides premature summaries, or nudges the conversation in particular directions, it may weaken the natural sense-making that good conversations rely on.
Human conversation often needs time. People need to hesitate, search for words, disagree, change their minds, and sit with uncertainty. A fluent AI response can easily make the conversation feel more settled than it really is.
There is also the danger that people begin to perform for the AI. If they know the conversation is being transcribed, summarised, or analyzed, they may speak more cautiously. They may avoid half-formed thoughts. They may be less willing to admit uncertainty or show vulnerability.
Conversational Leadership is about creating the conditions where people can think together. If AI starts directing the conversation, rather than supporting it, it can undermine that process.
The need for clear conversational norms
If AI is to join conversations well, we need some simple norms.
Before the conversation begins, people should know whether AI will be present and what role it will play. During the conversation, the group should be able to decide when to invite the AI in and when to keep it quiet. At the end, any AI-generated summary should be checked by the people who took part.
Some useful questions might be:
- Why are we bringing AI into this conversation?
- What role do we want it to play?
- Who can ask it to speak?
- Who can request that it be muted or removed?
- Will anything be recorded, stored, or shared?
- How will we check any summary or recommendation it produces?
These questions may seem simple, but they matter. Bringing AI into the room is not just a technical act. It is a social intervention.
A new participant in the conversational space
Seen positively, AI may become a new kind of participant in organizational conversations.
Not a decision-maker. Not a leader. Not a replacement for human judgement.
More like a background presence that can help us remember, connect ideas, retrieve knowledge, ask useful questions, and occasionally help us step back and see the conversation more clearly.
If used thoughtfully, this could strengthen conversational practice rather than replace it.
The real leadership challenge is not how to automate conversations. It is about designing conversational spaces where human and artificial intelligence work together in ways that improve how we think and act.
AI may join the conversation, but responsibility for it remains human.
Join one of my upcoming Knowledge Cafés or workshops and explore Conversational Leadership in practice.
Engage in thoughtful conversation, challenge assumptions, and connect with others working in knowledge, leadership, and organisational change.