Granola Keeps Winning on the "No Bot in the Room" Pitch

Granola continues to build momentum as the anti-bot meeting notetaker — it listens via your computer's own audio instead of sending a visible participant into Zoom, Meet, or Teams [1]. The pitch is simple: you jot down rough, imperfect notes during the call, and the AI fleshes them out afterward into structured notes with decisions and action items, using the transcript plus context like your calendar and role [2][3].

It's private by default, supports meeting-type templates, and now lets you chat or search across your entire meeting history — not just one transcript at a time [1]. It's available across macOS, Windows, iOS, Android, and even Apple Watch, which signals Granola is chasing the "notes follow you everywhere" use case rather than staying desktop-bound.

On X, the conversation around Granola has shifted from "cool notetaker" to "how do I stitch this into my actual workflow" — people are pairing it with Claude for deeper synthesis and asking for a genuinely all-in-one local alternative that handles transcription, summarization, and follow-through without duct tape.

Agents Are Starting to Act on Your Meeting Notes, Not Just Write Them

Two separate product threads — SmartJellyAI and Notion's Custom Agents — point at the same shift: meeting notes are becoming inputs for automated action, not just records. SmartJellyAI is a local Mac app that combines chat, dictation, and meeting support (prep, live assistance, post-call notes), then feeds those "memories" into other workflows so open loops and action items actually get tracked rather than buried [1]. The pitch is consolidation — one subscription instead of five overlapping tools.

Notion's AI Meeting Notes already transcribes and summarizes, but the bigger move is Custom Agents that auto-trigger after a meeting ends — updating trackers, posting recaps to Slack or email, even filing tickets, without a human initiating it [2][3]. This is currently a Business/Enterprise feature and runs through sub-processors like OpenAI and Anthropic, which is worth flagging for anyone thinking about data residency.

X users are already sharing setups that pull transcripts straight into Notion databases with auto-generated summaries and action items — the emphasis is increasingly on local storage and proactive follow-through, not just "nice summary, now go do the work yourself."

EU Pushes High-Risk AI Deadlines to December 2027 — But Transparency Rules Are Already Live

The EU's Digital Omnibus (Regulation (EU) 2026/1744, in force since July 27, 2026) has pushed back the toughest AI Act obligations. Standalone high-risk AI systems — think hiring tools, biometrics, education — now have until December 2, 2027 to comply, delayed from the original August 2026 deadline; AI embedded in products gets until August 2028 [1][2].

But don't read this as a free pass. Article 50 transparency obligations — disclosing when you're interacting with AI and labeling AI-generated content — are already in force as of August 2, 2026, with only limited grace for legacy systems [3]. For any Nordic or EU company running meeting transcription or AI-generated summaries, that means clear disclosure requirements are live now, even while the heavier risk-management and audit regime for high-risk use cases has more runway.

What This Means For Your Meetings

Today's news sketches out where meeting intelligence is headed on two fronts at once. On the capability side, Dessa shows that voice AI is getting closer to actually sounding and behaving like a person in conversation — interruptions, backchannels, emotional nuance — which matters enormously for tools that aim to sit in on meetings or act as a voice interface to your knowledge base. Meanwhile, Granola and the Notion/SmartJellyAI moves show the market converging on the same insight: raw transcripts are commodity now. The value is in what happens after — structured notes, synced trackers, auto-triggered follow-ups, and the ability to query your entire meeting history, not just today's call.

That last point is the one worth sitting with. A transcript of one meeting is a record. A searchable, speaker-identified, cross-meeting knowledge graph is institutional memory — it's the difference between "what did we decide last Tuesday" and "what has this client told us across the last six months, and who said what." The agent trend (Notion auto-triggering Slack recaps, SmartJellyAI feeding memories into other workflows) is the industry admitting that notes nobody acts on are just noise with better formatting.

The EU regulatory story matters here too, even in its delayed form. Transparency obligations are already live — if your meeting tool uses AI to summarize, label it as such. The heavier high-risk compliance wave is now 2027, but Nordic and EU teams building or buying meeting intelligence tools should treat this as a planning runway, not a reason to ignore data governance until the deadline.

Key takeaway: The real competition isn't who transcribes best anymore — it's who turns scattered meeting conversations into a living, queryable knowledge base that acts on your behalf, while staying honest about what's AI and what isn't.

Sources

https://www.theverge.com/ai-artificial-intelligence/1017906/jack-dorsey-block-dessa-ai-voice-speech
https://block.xyz/dessa
https://cellcog.ai/blog/block-dessa/
https://www.granola.ai/
https://www.nytimes.com/wirecutter/reviews/granola-meeting-notes-app-review/
https://www.granola.ai/blog/announcement
https://smartjelly.ai/
https://www.notion.com/releases
https://www.notion.com/help/ai-meeting-notes
https://digital-strategy.ec.europa.eu/en/policies/guidelines-ai-high-risk-systems
https://www.technology.org/2026/10/13/eu-ai-act-deadlines-after-the-digital-omnibus-what-applies-from-december-2026-to-august-2028/
https://www.goodwinlaw.com/en/insights/publications/2026/08/alerts-technology-dpc-eu-ai-act-transparency-obligations-now-in-force

Granola Keeps Winning on the "No Bot in the Room" Pitch