When Every Meeting Is Recorded, What Happens to Candor?AI meeting assistants are becoming part of everyday work. New research raises questions about what happens when those tools start evaluating the people in the room. I recently attended a meeting with almost as many AI agents as human participants. š Between meeting assistants, recording devices and smart glasses, Iām starting to wonder how much of our working lives are being documented, analyzed and stored somewhere. A few weeks ago, I received a meeting recap from a client who uses Read AI. I expected the usual summary, action items and next steps. What surprised me was the reportās assessment of individual participants. It included scores based on how much people spoke, their tone and how receptive others appeared to be to their contributions. Iāve gotten used to AI meeting summaries. I hadnāt expected a performance review from a meeting bot. As someone who has spent much of my career advising executives on how to communicate with employees, I found the experience unsettling. Iāve already seen leaders become more hesitant to have candid conversations with their teams or share information ahead of public announcements because theyāre concerned about recordings, leaks or comments being taken out of context. Now weāre introducing technology that can document those conversations and assign scores to the people having them. Iām concerned about what that could mean for the quality of communication inside organizations, especially when leaders are trying to build trust with employees.
When a Meeting Summary Becomes an EvaluationI wanted to understand what Read AI was measuring, so I looked into the companyās documentation. According to Read AI, its meeting scores combine measures of sentiment and engagement. Depending on the settings and location, the system analyzes verbal cues, speaking time and facial or other visual cues to estimate how participants are responding to a conversation. It also provides individual participant scores. Read AI describes these metrics as a way to assess meeting effectiveness. The company allows users to turn them off, which is worth knowing because many people may have no idea these features exist. The scoring raises some questions for me. How does a meeting assistant evaluate someone who deliberately stays quiet because theyāre listening carefully? What happens when an employee challenges an executiveās proposal and the conversation becomes uncomfortable? How does the software account for the person who asks an unpopular question that ultimately prevents the team from making a costly mistake? In communications, we spend a lot of time helping leaders understand their audiences. We know that a positive reaction doesnāt necessarily mean a message was effective. Silence can mean agreement, confusion, disagreement or simply that someone needs more time to think. And sometimes the most useful contribution to a meeting is the one that makes everyone uncomfortable. Iām skeptical that speaking time, vocal tone and visible reactions can tell us enough about the quality of someoneās contribution to justify a numerical score. Researchers Are Starting to Examine These SystemsA 2026 study by linguistics researcher Nicole Holliday, published in the International Journal of Applied Linguistics, examines a category of tools she calls socially prescriptive speech technologies. These systems evaluate how people communicate and offer feedback based on standards of supposedly desirable speech. Holliday studied Read AI and Zoom Revenue Accelerator and raised concerns about the accuracy and fairness of their evaluations. One particularly concerning result involved neurodivergent participants. Read AI negatively evaluated 58% of participants who self-identified as neurodivergent for language the system classified as non-inclusive. The study concluded that these systems can provide feedback that fails to account adequately for linguistic and social context. Thatās significant when we consider how differently people communicate. Someoneās speaking style can reflect their cultural background, personality, disability, professional experience or relationship with others in the meeting. A numerical score can make an interpretation appear more objective than it is. For communications leaders, this deserves attention. If these systems become part of employee development, performance conversations or leadership assessments, organizations need to understand what theyāre measuring and where the technology may be getting it wrong. What Happens When People Know Theyāre Being Watched?My other concern is what constant recording does to peopleās willingness to speak freely. Psychological safety has been an important concept in organizational research for decades. In a foundational 1999 study published in Administrative Science Quarterly, Harvard researcher Amy Edmondson examined 51 work teams and found that psychological safety was associated with learning behavior. The concept describes whether people believe they can take interpersonal risks, such as admitting mistakes, asking questions or challenging established thinking, without facing negative consequences. Think about how much of that behavior depends on the conditions in which conversations happen. An employee might feel comfortable questioning a decision in a small team meeting but hesitate if they know their comments will be recorded, summarized and available to people who werenāt present. A manager might avoid acknowledging uncertainty if an AI-generated recap could make an unfinished thought appear to be a formal position. And a senior leader might decide that sharing information early creates too much risk if thereās no control over where the recording or transcript could end up. These are plausible responses, and some reflect behavior Iāve observed in my own work. We still need more research to establish how frequently AI meeting assistants are causing people to change what they say. There is, however, a growing body of research on workplace monitoring that gives us reason to pay attention. In September 2025, the U.S. Government Accountability Office published a review of digital workplace surveillance based on 122 studies and interviews with stakeholders from 11 organizations. The report found that monitoring technologies can have positive and negative effects on employees. Some improve safety or help organizations manage work. Others can increase anxiety and stress, particularly depending on how employers collect and use the information. The GAO also warned that flawed performance measures or uses that werenāt originally intended could create negative employment consequences. Meeting assistants arenāt identical to the employee monitoring systems examined in that report. But the underlying concern is relevant: when organizations collect information about peopleās behavior, how that information is interpreted and used can affect employees. The Leadership Communication ProblemThis is the part I keep coming back to. For years, internal communications professionals have encouraged leaders to communicate earlier, share more context and be willing to have difficult conversations with employees. We encourage executives to explain the thinking behind decisions, acknowledge uncertainty and make themselves available for questions. Those practices require a certain amount of trust. Leaders need to believe they can have a conversation without every sentence becoming a permanent corporate record that might circulate outside its original context. Employees need to believe they can ask difficult questions without those questions being used to judge their attitudes, engagement or performance. Iāve worked with leaders who are already cautious about sharing information before a public announcement because they worry about leaks. I can see how the possibility of a recorded transcript or AI-generated interpretation could make them even more reluctant. Consider a leader discussing a potential restructuring before the details are finalized. They might want to explain what the organization is considering and give employees an opportunity to ask questions. But if that discussion is recorded and the summary is circulated without the surrounding context, tentative ideas could be mistaken for final decisions. The safer choice for that leader may be to wait until everything is finalized and approved. From an internal communications perspective, thatās a concern. Employees often want to understand how decisions are being made, especially when those decisions affect their jobs. Delaying communication can leave people with fewer opportunities to ask questions and contribute. I worry that weāre making it harder for leaders to practice the kind of communication weāve spent years encouraging. Thereās an employee dimension to this, too. Someone who disagrees with a manager may be willing to raise a concern during a live conversation but hesitate if the exchange becomes part of a searchable record. An employee who is already unsure whether itās safe to speak up may become even more cautious when an AI system is assessing their tone or apparent engagement. We should be asking whether these tools are changing the conditions under which people feel comfortable speaking honestly. We Need Better Rules for AI in MeetingsI use AI regularly, and I understand why meeting assistants have become popular. Theyāre useful for capturing decisions, documenting next steps and helping people catch up on conversations they couldnāt attend. For some employees, transcripts and summaries can also make meetings more accessible. Iād be reluctant to give up those benefits. But organizations need to be more deliberate about how they introduce these tools, especially as meeting assistants expand from documentation into behavioral analysis. I think communications leaders should be involved in at least 4 decisions. First, which meetings should be recorded? A routine project update and a sensitive conversation about organizational change have different purposes. Teams should have the ability to hold certain conversations without recording or automated analysis. Second, what are meeting analytics allowed to measure and influence? If a tool generates scores for engagement, sentiment or individual participation, employees should understand how those scores are calculated, what their limitations are and whether anyone can use them in performance decisions. My preference would be to keep speculative behavioral scores out of employee evaluations. Third, who has access to the information? Meeting recordings, transcripts and summaries can contain sensitive business information and personal comments. Organizations need clear rules for access, retention, sharing and deletion, including what happens when an outside participantās meeting assistant joins the call. Fourth, how do we preserve opportunities for candid conversation? Leaders and employees need settings where they can ask questions, challenge assumptions and discuss unfinished ideas. That may mean establishing meeting norms that explicitly allow participants to turn off recording or choose a different format for sensitive discussions. These are communications decisions as much as technology and legal decisions. They affect how people experience leadership, whether they trust the organization and how willing they are to contribute. What Kind of Conversations Do We Want to Have?I suspect weāre still early in understanding how AI meeting assistants will affect workplace behavior. The technology is developing quickly, and many organizations are adopting it because the immediate benefits are easy to see. A summary arrives within minutes. Action items are organized. Nobody has to volunteer to take notes. The consequences for communication are harder to measure. How do we know whether someone withheld an important concern because they were uncomfortable being recorded? How do we account for an executive who chose to deliver a carefully scripted update rather than invite an open discussion? Or the employee who decided an unpopular opinion wasnāt worth having permanently attached to their name? Those conversations that never happen wonāt appear in a meeting recap. For internal communications professionals, I think this deserves much more attention. Weāre responsible for helping organizations create conditions where people can exchange information, ask questions and build trust. The technologies we introduce into those conversations can influence whether that happens. I want the convenience of AI meeting assistants. I also want workplaces where people can disagree, admit uncertainty and have conversations that arenāt automatically evaluated by software. Iām curious how other communications leaders are approaching this. Are you seeing recording and meeting analytics change how executives and employees communicate? Have you established guidelines for when AI assistants can participate in meetings? Iād especially like to hear from people who have found a good balance between useful documentation and the need for candid conversations. |
Comms FutureLab is a media platform for communications and marketing leaders who are endlessly curious about the future of our profession. Through thoughtful analysis of AI, media, leadership, trust and organizational change, we help today's leaders prepare for tomorrow's challenges and opportunities.
We're Changing How We Use AI. Are We Changing How We Manage It? AI is becoming more involved in how we work. A recent announcement from OpenAI raises questions about what happens when the technology we've come to depend on changes. If you've spent hours building a custom GPT, you might want to pay attention to December 11, 2026. That's the date OpenAI plans to retire custom GPTs and move users toward plugins. For anyone who has invested time creating a specialized AI assistant, this is a...
Your Embargo Needs an Internal Comms Plan Iām working on a project with an organization that has a longstanding practice: employees hear company news 24 hours before itās announced publicly. I understand the reasoning. Employees shouldnāt routinely learn important news about their company from a news alert, LinkedIn post or reporter. But 24 hours can be a long time when youāre also working with media under embargo. An embargo gives reporters access to information before an announcement with...
AI Has a Communications Problem Sam Altman has been making the rounds, and in a recent interview with Bloomberg, he acknowledged that the AI industry has done a poor job communicating the benefits of the technology. Itās not often you hear ācommunications challengeā in a Bloomberg cover story, so of course I was instantly invested. Altman talked about the fear surrounding AI and the need to give people a greater sense that it will increase their agency rather than take it away. I agree with...