The Quiet Defection
Granola launched without much fanfare – no splashy product launch event, no celebrity investor announcements, no viral Twitter thread from a founder. It was a Mac-only AI notepad built around a simple premise: you type rough, sparse notes during a meeting, and the app uses those notes alongside audio to generate something actually useful afterward. That’s it. No bot joining your call. No robotic voice announcing it’s recording. No one on the other end knowing anything is happening at all.
That last part turned out to matter more than anyone anticipated.
Otter.ai built its business on being visible. The little Otter bot joins your Zoom, announces its presence, and dutifully transcribes everything. For years, that was fine – people tolerated it the same way they tolerate a conference room speakerphone. But tolerance is not loyalty, and a growing number of power users – the consultants, product managers, VCs, and executives who live in back-to-back meetings – have started migrating to Granola with a speed that suggests they were waiting for exactly this kind of alternative.
What Granola Actually Does Differently
The core mechanic sounds almost too simple. Granola runs locally on your Mac and listens to your meeting audio – system audio and microphone – without injecting anything into the call itself. You jot a few words as things come up: “Q3 budget – push back from Sarah”, “timeline slipping – ask about dependencies.” After the meeting ends, Granola combines those notes with the full transcript to produce structured, context-aware summaries that actually reflect what mattered to you, not just what was said. The difference between a transcript and a useful summary has always been the hardest problem in this category. Granola’s approach to solving it – using your own sparse annotations as a signal – is genuinely clever because it offloads the curation work to the only person who actually knows what was important: you.
Otter.ai’s strength was always volume. It captures everything, and that comprehensiveness was the product. But volume creates its own problem – long, undifferentiated transcripts that still require someone to sit down and extract the decisions, action items, and follow-ups buried inside them. Power users eventually hit a wall where Otter was generating more text than it was saving time. The app became another inbox. Granola solves this not by being smarter about summarization in some abstract AI sense, but by changing what goes into the summary in the first place.
There’s also the privacy angle, which has become louder as AI meeting tools proliferate. Granola never joins calls as a participant, which means it sidesteps a genuinely uncomfortable social dynamic. When a bot announces itself at the start of a client call or a delicate internal discussion, it changes the conversation. People self-censor. Candor shrinks. Granola’s invisible recording model removes that friction entirely – though it does raise its own question about whether the people on the other end have consented to being recorded at all, a legal distinction that varies significantly by jurisdiction.
Why Otter’s Power Users Are the Ones Leaving
Churn in productivity software tends to follow a predictable pattern: casual users leave first because they never built deep habits, and power users stay longest because they’ve invested in workflows, integrations, and team adoption. Granola appears to be breaking that pattern. The defectors are not people who used Otter occasionally and forgot to cancel their subscription. They are the people who were Otter’s best argument for the product – the ones with hundreds of recorded meetings, refined templates, and colleagues who had grown accustomed to receiving Otter summaries after every sync. That demographic shift matters because power users are also the ones who drive word-of-mouth adoption inside organizations, who push for team licenses, and who show up in case studies.
The workflow fit for Granola skews toward individual contributors and senior professionals who operate with a lot of autonomy over their own time. A VC doing eight calls a day has very different needs than a customer support team trying to log interactions at scale. Otter’s enterprise features – team workspaces, shared libraries, Salesforce integrations – still make it the more defensible choice for structured organizational use cases. Granola is not trying to replace that yet. What it is replacing is Otter’s hold on the individual professional who pays out of pocket, uses the tool for their own benefit, and doesn’t need a CRM sync. That segment was quietly Otter’s most reliable growth engine.
The Mac-only limitation is the obvious counterargument. A significant portion of knowledge workers use Windows, and Granola simply does not exist for them. This is either a temporary constraint the company will address or a deliberate signal about its intended market – the Apple-ecosystem professional who is already predisposed toward polished, opinionated software. Tools like Warp’s AI terminal have followed a similar trajectory, building devoted Mac-first user bases that create enough momentum to fund broader platform expansion later. Whether Granola follows that path or stays narrow by design remains open.
The Window Is Narrow
Otter.ai is not standing still. The company has been adding AI summary features, action item extraction, and integrations at a steady clip. But there’s a structural problem: adding AI on top of a bot-joins-the-call model does not fix the bot-joins-the-call model. The user experience is still defined by that visible, announced presence, and no amount of post-meeting AI polish changes what happens at the start of the recording. Granola’s advantage is architectural, not incremental, which means Otter can’t fully replicate it without rebuilding from a different foundation. The real risk for Granola is not Otter catching up – it’s Notion, which already has millions of users storing meeting notes, or Apple itself, which controls the system audio layer Granola depends on and has every incentive to build this directly into macOS.
