What are AI proxy meetings?

AI proxy meetings are video calls where one or more humans send an AI bot — a notetaker, summarizer, or agentic AI — to attend in their place rather than joining themselves. The bot records, transcribes, and increasingly decides what context the absent human sees afterward. In 2026, AI proxy meetings have become a default behavior in many US teams, driven by meeting overload, cross-platform AI notetakers, and the leader-IC AI adoption gap.

How are AI proxy meetings different from AI notetakers?

AI notetakers were originally a transcription tool that supplemented human attendance. AI proxy meetings are about *attendance by delegation* — the AI is no longer supplementing a human in the room, it is replacing one. The same Otter, Granola, Fathom, or Zoom AI Companion 3.0 tool that started as a recorder is now used to skip the meeting entirely, with the human relying solely on the AI's summary. That shift changes everything about privacy, trust, and decision quality.

Why are clients refusing meetings with AI proxies?

Because AI proxy meetings break trust in three ways: (1) the absent human is not actually engaged, (2) the bot's transcript can leak outside the meeting envelope, and (3) the bot's summary becomes the \"ground truth\" the absent decision-maker acts on. Sales prospects, partnership counterparties, and confidential investors are increasingly asking for bot-free meetings as a baseline trust signal in 2026.

What's the alternative to AI proxy meetings?

The emerging model is canvas-first preparation, AI that augments humans actually in the room (rather than replacing absent ones), and explicit bot policies that turn AI proxy meetings from a quiet liability into a managed practice. Coommit is one of the platforms built around this model — a video meeting tool with an interactive canvas and contextual AI that strengthens the humans who show up rather than replacing the ones who left.

AI Proxy Meetings: When Your Bot Shows Up Without You

Q: Are AI proxy meetings legal?

It depends on jurisdiction and meeting type. Many US states require explicit consent from all participants to record a call (two-party consent states like California). The NYC Bar Association's Formal Opinion 2025-6 warns that AI proxy meetings can void attorney-client privilege when the bot's transcript is redistributed. HIPAA and GDPR add further restrictions. Teams operating AI proxy meetings without a documented consent flow and a clear bot policy are taking on legal risk most CIOs have not priced in.

# AI Proxy Meetings: When Your Bot Shows Up Without You

Last summer, a Portland engineer named Clifton Sellers joined a Zoom call and started counting. Six humans on the grid. Then he opened the participant list. Ten AI notetakers were already there, sent by colleagues who had something better to do. The bots outnumbered the people. By May 2026, that scene is no longer a curiosity — it is the default. Fortune reported on May 20 that managing partners are watching clients walk out of meetings the moment a notetaker bot dials in. And the practice now has a name: AI proxy meetings.

AI proxy meetings — calls where one or more humans send a bot instead of attending themselves — are reshaping how American teams work in 2026. 47% of enterprise AI users say they have already based a major business decision on hallucinated content. Manager engagement just dropped from 27% to 22% in a single year. The math is brutal. The trust is gone. And nobody is asking the obvious question: what exactly are AI proxy meetings replacing, and what are they breaking on the way?

This deep-dive unpacks what AI proxy meetings actually are, the four forces driving their explosion in 2026, the four hidden costs nobody is pricing in, and what comes next when the bot is the meeting and the meeting is the bot.

What AI proxy meetings are (and why "AI notetaker" no longer covers it)

The phrase AI notetaker was coined for tools like Otter, Fathom, Fireflies, and Granola that join a call, transcribe it, and email a summary. That definition fit in 2023. It does not fit 2026.

Today, AI proxy meetings describe a broader pattern: a human delegates their *presence* — not just their notes — to an AI. The bot does more than listen. It records, summarizes, decides what gets flagged to the absent human, and increasingly speaks back into the room. With Zoom AI Companion 3.0 rolling agentic workflows and a "My Notes" notetaker that now joins Microsoft Teams and Google Meet calls as a third-party bot, the AI proxy meeting has gone cross-platform. With Google Meet's new "Decisions" capture auto-logging conclusions whether or not the decision-maker is on the call, the proxy is also doing governance.

So AI proxy meetings are not really about transcripts anymore. They are about *attendance by delegation*. And once you frame them that way, the implications get a lot more interesting — and a lot more uncomfortable.

The distinction matters because attendance by delegation is a different product category than transcription. It implies the bot has authority. It implies decisions made in the room bind the absent. It implies the absent human will accept the bot's filtered summary as ground truth. None of those assumptions are tested. All of them are now operational in millions of US teams.

The four forces driving AI proxy meetings in 2026

AI proxy meetings did not arrive by accident. Four forces conspired to make them inevitable.

Force 1 — A meeting overload nobody ever fixed

US knowledge workers spend about 11.3 hours per week in meetings — 28% of the workweek — and 72% of those meetings are rated ineffective. We added Zoom. We added Slack. We added asynchronous Loom videos. The meeting count went *up*, not down. Sending an AI proxy to the meeting is the rational individual response to an irrational collective system. If the only options are "skip the meeting and get blamed" or "attend the meeting and lose two hours of focus time," sending a bot looks like a third way. It isn't — but it feels like one. This is also why our piece on no-meeting days for remote teams went viral last year. The pain is universal.

Force 2 — AI notetakers that became status symbols

Somewhere between 2024 and 2026, having a notetaker bot in your seat became a flex. It signaled "I'm too important to be present" the way a BlackBerry once signaled "I'm too important to wait." Vendors leaned into it. Granola, Fathom, Otter, and every Zoom/Google/Microsoft-native equivalent now ship with one-click "send me even when I can't make it" features. The result is exactly what you'd expect: AI proxy meetings have become the corporate equivalent of the participation trophy. Everyone gets one. Nobody is actually there.

Force 3 — The leader-IC adoption gap

Asana's State of AI at Work found that 52% of executives use AI weekly versus only 36% of the broader workforce and 25% of individual contributors. Executives are the heaviest senders of AI proxies. They are also the loudest about "alignment." That gap creates a brittle dynamic: the leader sends a bot, the team adapts to a meeting attended by a transcript, and decisions get retroactively rationalized through the bot's summary. We unpacked this asymmetry in the team AI adoption gap. AI proxy meetings are where that gap becomes a productivity tax on everyone else.

Force 4 — Cross-platform agentic AI that never sleeps

In May 2026, two announcements turned AI proxy meetings into infrastructure. Microsoft's 2026 Work Trend Index confirmed that "agentic" AI is now embedded in M365 and Teams by default. Google launched Gemini Enterprise with an Agent Runtime that maintains memory across sessions — meaning a single AI proxy can attend dozens of meetings over several days and synthesize them into a single brief. The proxy is not a meeting attendee anymore. It is a persistent identity that participates in every meeting on your calendar, with or without your input. That is a fundamentally different product than an Otter recording.

The four hidden costs of AI proxy meetings nobody is pricing in

The marketing pitch for AI proxy meetings is seductive: reclaim your calendar, never miss context, scale your presence. The reality is four hidden costs that show up later, usually after the damage is done.

Cost 1 — Privilege and confidentiality collapse

In the legal world, this is already a crisis. A managing partner quoted in Fortune on May 20 said it bluntly: "The amount of times I have seen clients lose privilege because an AI notetaker disseminates notes from a confidential meeting to a group of people without a lawyer in copy." The NYC Bar Association issued Formal Opinion 2025-6 warning that AI proxy meetings can void attorney-client privilege when the bot redistributes content outside the legal envelope. The same principle applies to HIPAA, GDPR, and any board-level discussion governed by NDAs. AI proxy meetings turn a single oversight into a discoverable transcript living in three different SaaS vendors. We covered the compliance dimension in banning AI notetakers, and the trend has only accelerated.

Cost 2 — Decisions based on hallucinated context

The Hacker News reported in May 2026 that 47% of enterprise AI users have based a major business decision on hallucinated AI content. AI proxy meetings are the perfect vector. The absent human reads the bot's summary, takes it as ground truth, and acts. There is no friction layer. No double-check. The summary becomes the meeting. When the summary is wrong — and at current frontier hallucination rates of 20-30% on complex multi-turn tasks, it sometimes is — the decision is wrong. The bot is not accountable. The absent human shrugs and says "the AI got it wrong." The team that was actually in the room eats the consequence. We saw similar dynamics in meeting AI amnesia — the gap between what the bot captures and what humans remember is now a measurable performance tax.

Cost 3 — Clients refuse meetings with bots in the room

The blowback is starting to show up in revenue. The UMEVO "Bot Backlash" report tracked a sharp rise in US clients explicitly requesting "no bots, no recording" before agreeing to a meeting. Sales calls, partnership discussions, and high-trust founder conversations are increasingly bot-free zones. If your sales team sends a Granola or Fathom proxy to a discovery call, the prospect may quietly mark you as unsafe and never tell you why. AI proxy meetings are not free. They cost trust. And trust has a longer half-life than any AI summary.

Cost 4 — The meeting loses its actual purpose

Meetings are not transcripts. They are mechanisms for alignment, relationship building, and decision-making — the messy social functions that make organizations work. Gallup's 2026 State of the Global Workplace found US employee engagement at 31% (down from 36% in 2020) and manager engagement collapsing from 27% to 22% in a single year. The correlation is not subtle. When humans stop showing up to each other's meetings — when AI proxy meetings become normal — the relational substrate that makes teams function quietly erodes. Bots cannot rebuild it. We dug into the broader signal in context fatigue: people are not just exhausted by *more* meetings, they are exhausted by meetings that no longer feel like meetings.

Why "more AI" is the wrong fix for AI proxy meetings

The first instinct of most CIOs in 2026 is to fix AI proxy meetings with more AI. Better summarization. Smarter agents. Multi-modal context. BCG's 2026 AI Radar reports that 60% of companies see "hardly any material value" from AI despite doubling spend to 1.7% of revenue. The reason is structural: layering more AI on a broken meeting culture does not fix the meetings. It accelerates the proxy. It strengthens the bot. It widens the leader-IC adoption gap. And it makes the four hidden costs worse, not better.

The Pew Research Center finds that only 21% of US workers say at least some of their work is done with AI — just 2% say "all or most". The vendor narrative says AI is everywhere. The worker reality says AI is a thin layer that mostly summarizes meetings nobody wanted to attend. AI proxy meetings sit on top of that thin layer and pretend it is a foundation. It is not.

What comes after AI proxy meetings

The honest answer is that the meeting itself needs to change before AI proxy meetings stop being rational. Three shifts are already starting in 2026.

The first shift is canvas-first preparation. Instead of a synchronous call with a bot transcribing, teams pre-build the meeting's content — proposal, diagram, decision criteria — on a shared canvas before the call starts. The synchronous time is then spent on the parts that genuinely require human presence: disagreement, judgment, and commitment. The canvas becomes the persistent artifact. The recording, if any, becomes a redundant safety net rather than the primary deliverable. This is the model we built Coommit around, and it is increasingly the model adopted by teams that have hit the limits of AI proxy meetings.

The second shift is AI that augments present humans rather than replacing absent ones. The version of meeting AI that works is the one that listens to the humans who are actually in the room and surfaces context they would otherwise miss — past decisions, related canvas artifacts, evidence supporting or contradicting a claim. This is fundamentally different from an AI proxy meeting, where the AI represents an absent human. One model strengthens the room. The other replaces it. We explored the distinction in AI meeting agent vs notetaker, and the gap is widening fast.

The third shift is explicit bot policy as a trust signal. The companies winning in 2026 are publishing public meeting bot policies — what bots are allowed, when, by whom, with what consent flow — and using them as a trust signal to clients, partners, and employees. A clear policy turns AI proxy meetings from a quiet liability into a managed practice. A clean AI meeting bot policy is now table stakes for any team that takes confidentiality, client trust, or simple meeting hygiene seriously.

Together these three shifts point to a post-proxy future where AI is in the meeting, but the meeting is for humans. Where the bot helps, but the bot does not attend on anyone's behalf. Where the persistent artifact is the canvas, not the transcript. And where the meeting count drops because the meetings that survive are the ones that needed humans in the first place.

The bottom line

AI proxy meetings are a symptom, not a strategy. They are the rational individual response to a meeting culture that broke and a leader-IC adoption gap that nobody fixed. They are also costing companies privilege, trust, decision quality, and engagement. The fix is not better AI proxies. The fix is fewer meetings, better preparation, and AI that strengthens the humans in the room rather than replacing the ones who left. Coommit was built for that future — a video meeting tool with a real-time canvas and contextual AI that augments the people actually showing up. If you are tired of being one of six humans on a call with ten bots, you already know which side of the proxy line you want to be on.