A reflection on why the growing abundance of AI summarisation makes preserving genuine disagreement a more valuable leadership skill, not a less relevant one.
Disagreement is often untidy. People repeat themselves. Important concerns appear beside minor ones. Some participants speak directly, while others communicate uncertainty through silence, hesitation or carefully qualified language.
A good summary helps a group move forward. It identifies shared conclusions, clarifies unresolved questions and separates the central issue from the noise around it. Artificial intelligence is becoming very good at producing this kind of summary — analysing a long meeting and returning a short list of decisions, themes and next steps.
This is valuable. But every summary is also an act of interpretation. Someone — or something — decides which differences matter, which comments belong together, and which qualifications can be removed without changing the meaning. When AI performs this work, a group may receive the appearance of consensus before genuine consensus exists.
A meeting may contain three different forms of silence: one person agrees, another does not understand, a third disagrees but does not feel able to challenge someone senior. The transcript may record silence in exactly the same way. An AI summary might state that "the group broadly supported the proposal," even though the social reality was more complicated.
This possibility becomes especially important in cultures and organisations where disagreement is not always expressed directly. My doctoral research examined characteristics such as power distance, consultation, respect for leadership and the ability of junior members to express different opinions within technology-mediated communities. One practical implication was that community design should create opportunities for consultation rather than assuming that the absence of visible opposition represents agreement.
The intelligent age gives this earlier question new significance. AI does not merely transmit the discussion — it may become the official interpreter of what the discussion meant. Once a summary enters the meeting record, it can acquire authority. People who remember the conversation differently may hesitate to challenge a document that appears neutral and comprehensive.
The summary then shapes the next discussion. An unresolved concern disappears from the record. A minority position becomes a minor footnote. A conditional agreement is remembered as approval. Nothing was deliberately falsified — the complexity was compressed until the remaining narrative became more certain than the people who created it.
This is synthetic consensus: agreement produced partly through the process of summarisation rather than through a genuine meeting of understanding.
The risk extends beyond meetings. AI may summarise employee sentiment, citizen feedback, customer complaints or public consultation. A thousand different experiences may become five themes on an executive dashboard. The themes may be accurate at a broad level while removing precisely the exceptions that leaders most need to understand. A repeated complaint becomes visible; a rare but severe consequence may disappear because it occurred too infrequently to form a pattern.
Responsible use of AI summaries therefore requires more than checking grammatical accuracy. Leaders should ask: What disagreement was preserved? Which voices were underrepresented? What uncertainty did the summary remove? Does "most participants" conceal a group that will carry a disproportionate consequence?
The objective should not be to make summaries longer and less useful. It should be to preserve the distinction between common ground and unresolved difference. An intelligent system might label levels of confidence, identify contradictory views, and show where apparently similar statements were made for different reasons. Meeting owners might invite participants to correct the summary before it becomes the official record. Leaders might deliberately ask for the strongest minority view rather than only the dominant conclusion.
The ability to summarise will become increasingly abundant. The willingness to remain with disagreement may become more valuable. Human groups do not always need immediate harmony. Sometimes they need a more accurate understanding of why agreement has not yet been reached.
A clear summary can help a group move forward — but clarity becomes dangerous when it makes unresolved disagreement disappear.