Every productivity category eventually gets an AI layer. Meetings are no different — but the way AI is integrating into meeting rooms is less uniform than the marketing suggests. There are two genuinely different approaches, and they produce very different outcomes for the teams using them.
The first wave: AI as a visitor
The earliest AI meeting tools were add-ons: a bot you invited to your call. It joined as a visible participant, announced itself, recorded the audio, and sent you a transcript and summary afterward.
The appeal was obvious — it worked across any video platform you already used. The friction was also obvious: you had to remember to invite it. Someone occasionally forgot. Sensitive calls made attendees uncomfortable with an unrecognized participant recording everything. And the summary arrived via a third-party service that now held your meeting data.
This model is still dominant — Otter.ai, Fireflies.ai, and dozens of similar tools operate this way. They're useful. But they solve the transcript problem while introducing new ones around reliability, data custody and participant trust.
The second wave: AI as infrastructure
The more interesting shift is AI that doesn't arrive as a visitor — it's already in the room, built into the meeting platform itself.
This changes the dynamic in ways that matter. There's nothing to invite, so coverage is consistent. There's no separate participant making people pause before saying something sensitive. The transcript and recap aren't held by a third-party service — they're part of the meeting record, subject to whatever data policy applies to the meeting itself.
Platforms built on this model — MeetOye is one example — treat the AI layer as the default infrastructure of every call rather than an optional add-on. Oya, MeetOye's built-in assistant, transcribes and translates in real time, then emails a recap with decisions and action items to every attendee automatically when the call ends.
The translation gap
One specific area where AI is making a measurable difference: live translation.
The previous approach to multilingual meetings was simultaneous interpretation — an expensive human service reserved for conferences and executive calls. The tier below that was nothing: participants in a meeting they only half-understood, relying on the dominant-language speaker to summarize.
AI translation has moved this from a conference luxury to a per-meeting default. The more sophisticated implementations don't just show one shared translated caption for the room — they give each participant their own caption stream in the language they chose, regardless of what language the speaker is using. That distinction matters more than it looks: one shared caption language still requires participants to read in a second language; per-participant translation lets each person choose their own.
What AI still doesn't do well in meetings
Honest assessment:
Follow-up enforcement. AI can write an action item. It cannot make anyone do it. The accountability layer is still social and organizational.
Judgment about what to escalate. A transcript of a tense conversation and a structured summary of a productive one look similar to a language model. The nuance of what matters is still human.
Real-time facilitation. AI can suggest an agenda and flag when time is running short. It cannot redirect a derailed conversation or manage the room dynamics of a difficult discussion.
The practical takeaway
AI in meetings is most valuable when it handles the mechanical cognitive work — capturing, labelling, summarizing, translating — so that human attention can stay on the judgment-intensive parts: listening, deciding, responding.
The teams getting the most out of it are treating the AI layer as infrastructure: always on, consistently applied, not something that requires anyone to remember to set up. That shift from optional add-on to default behavior is where the actual productivity gain lives.
Author bio:
The MeetOye Team covers AI and the future of work. MeetOye (meetoye.com) is an AI-native video meeting platform with Oya, a built-in assistant that transcribes, translates per participant, and recaps every call automatically.
Top comments (0)