Juliette, our meeting intelligence platform, has transcribed and analyzed over 1,534 meetings as of this writing. It started as a productivity tool — join meetings, take notes, extract action items. Simple enough. What we did not expect was that the aggregate data would become a mirror that shows organizations how they actually operate, as opposed to how they think they operate.
The first thing the data showed us: about 35% of meetings are recurring meetings where the same information is repeated because the attendees from the last meeting did not follow through on their action items. The meeting exists to re-establish shared context that was lost because nothing happened between meetings. This is not a meeting problem. It is an accountability problem. But organizations keep scheduling more meetings instead of fixing the accountability gap.
The second insight was about decision-making. In the meetings we analyzed, less than 20% contained a clearly articulated decision. The rest were discussions, updates, or explorations that ended without a resolution. Many of these non-decision meetings spawned follow-up meetings to continue the discussion. We saw chains of four or five meetings about the same topic, none of which produced a decision. The AI summary of each meeting essentially said: no decision was reached and a follow-up has been scheduled.
The paradox of meeting intelligence is this: the better the tool captures what happened in a meeting, the more obvious it becomes that the meeting should not have happened. When you read a crisp, accurate summary of a 60-minute meeting and realize the entire content could fit in a three-paragraph email, it is a confronting experience. We have had clients review their meeting analytics and immediately cancel 30% of their recurring meetings.
Action item tracking is where Juliette creates the most tangible value. The system extracts commitments from meeting transcripts — things like I will send the report by Friday or the engineering team will investigate the bug. It tracks whether those commitments were completed by the next meeting. The completion rate across our user base is startling: about 40%. Sixty percent of commitments made in meetings are never fulfilled. People say yes in the moment and then get overwhelmed by other priorities.
We discovered that the most productive teams have the fewest meetings. This is not a new observation, but having data to prove it is powerful. Teams in the top quartile of output per person average 4 hours of meetings per week. Teams in the bottom quartile average 12 hours. The correlation is not perfect — there are high-meeting, high-output teams — but the trend is unmistakable.
One of the most interesting patterns is what I call the shadow decision network. The official org chart says that decisions flow through certain channels. The meeting data shows that decisions are actually made in small, informal meetings between specific individuals, and then ratified in the official meetings. Identifying these shadow decision-makers is incredibly valuable for change management. If you want to move an organization, find the people who are in the small rooms where things actually get decided.
The sentiment analysis features have been surprisingly useful for HR and leadership teams. Tracking how the emotional tone of team meetings changes over time can signal problems before they become crises. We have seen cases where meeting sentiment declined steadily over eight weeks before a team experienced significant turnover. The data was there. Nobody was watching it.
Privacy is obviously a concern, and we take it seriously. All transcription data is owned by the client. We do not use client meeting data to train models for other clients. Access is controlled by the client's own permissions. But even with these safeguards, some organizations are uncomfortable with the idea of AI listening to their meetings. Interestingly, the resistance usually comes from leadership, not from the team members who are tired of taking notes manually.
The meta-lesson from Juliette is that organizations have a massive blind spot about how they spend their time. They track financial expenditure meticulously. They track time spent in meetings not at all, or only at a superficial level. Meeting intelligence tools do not just make meetings better. They reveal the organizational dysfunction that meetings are papering over. That revelation is uncomfortable, but it is the first step toward actually fixing things.