
Markus Kellermann Founder & CEO
Granola AI Review 2026: Workflow, Pricing & Real Limits

Markus Kellermann Founder & CEO
Granola tested for 3 months across 60+ meetings. Notes on the workflow, the 25-note plan, the Google Workspace dependency, and the gaps.
Granola Is for People Who Type During Meetings
Three months back I started keeping Granola open on my second monitor for every client call. Not to test it for this review yet, honestly. My co-founder Iván had been using it for two weeks and kept sending me his enhanced notes from our 1:1s with the subject line "ok this is good actually," and I wanted to see what he meant.
He was right and wrong. The first thing I had to learn about Granola is that you don't use it the way you'd use Otter. It's not really a recorder. You take rough notes during the call, and the AI tidies them up afterward. Once you start using it that way, the rest of the product makes sense.
That's the review in three sentences, if you want the short version. If you take notes during meetings, Granola will probably make your week better. If you don't, almost nothing in this post will apply.
The Granola you're evaluating in 2026
Worth knowing what you're buying. Granola was founded in 2023 by Chris Pedregal (ex-Gmail PM at Google, previously founder of Socratic, which Google acquired in 2018) and Sam Stephenson, headquartered in London. They closed a $125M Series C in March 2026 at a $1.5B valuation, officially making them the latest UK unicorn in the productivity space.
The strategic direction matters too. Granola is no longer pitching itself as a notetaker. The current framing is "enterprise AI context layer," meaning Granola wants to be the central source of your team's meeting data, with APIs and MCP integration that pipe that data into Claude, ChatGPT, and every other AI tool your team uses. The MCP integration launched in February 2026. Spaces (team workspaces with granular access controls) launched alongside the Series C. The iOS app, which now also transcribes phone calls, launched the same quarter.
In other words: when you're evaluating Granola today, you're evaluating a unicorn pivoting from prosumer notetaker to enterprise data infrastructure. That's a different review from what existed eighteen months ago, and the parts of the product that feel rough at the edges are exactly the parts that the company hasn't optimised for yet under the new direction. For context, Convo is taking the opposite bet at much smaller scale, focused on real-time help during conversations rather than enterprise context layers built around them.
The write-then-enhance loop
You open the app before a call. Blank notes view. Type whatever comes to mind during the meeting. Rough stuff. "$ pushback on Q3 number." "Needs SOC2 by Sept." "Iván says he can fix the export bug." After the call, hit Enhance Notes. The AI takes your bullets, pulls in the silent transcript it captured while you were talking, and gives you back a structured summary, action items, decisions, and a few direct quotes from the call.
Output is good. Maybe 90% of what I cared about, with 10% needing minor edits. Better than I expected when I started. Worse than the Granola homepage suggests, which is fine. The homepage of every AI product oversells.
Two things this gets right that nobody else in the category does. One of them is invisible, which is probably why nobody else has copied it yet.
The visible one first. Granola pulls audio from your machine's system output instead of joining the call as a participant. Nothing shows up in the Zoom or Meet attendee list. I asked three different clients (one consultant, one healthcare admin, one engineering manager at a manufacturing company) whether they'd noticed anything weird on our calls. Nobody had. The visible-bot problem that Otter, Fireflies, and Fathom all share just isn't there. Worth a lot on relationship-heavy calls. Worth basically nothing on internal team syncs where everyone knows you're recording anyway.
The invisible one is the assumption baked into the workflow. Every AI meeting tool I've used in the last two years has been promising the same thing: you can be fully present, the bot will handle everything. This is a lie. The moment you stop actively processing the conversation, you stop processing it. Bots that summarize for you turn you into a passive listener, and passive listeners ask worse questions. Granola makes you keep listening because you have to keep typing. Cleaner output, sharper meetings, mildly tired wrists.
Granola ships 29 meeting templates that shape the enhanced output for different meeting types: discovery call, 1:1, planning, pipeline review, customer onboarding, sales debrief. Each one biases the AI toward what matters for that meeting shape. The Pipeline Review template emphasises deal progression and risk. The 1:1 template emphasises blockers and personal commitments. You can also build your own. The templates aren't revolutionary, but they're a real improvement over generic AI summaries. Convo ships a smaller set of templates focused on live coaching during the call rather than post-call summarisation, which is the structural difference these tools never converge on.
Looking for help during the call, not just better notes after? Convo does the bot-free capture Granola pioneered, plus real-time suggestions, context from past calls, and objection responses while the conversation is still happening. Start a 7-day free trial.
Three things it costs you
Three problems that hit by month two.
1. Granola assumes you're on Google Workspace
The product is built around Google Workspace and it shows. The calendar integration that triggers transcripts uses Google Calendar. Login is a Google account. If your work email is on Microsoft 365, Zoho, or any custom domain that isn't connected to Google Workspace, things get flaky in ways that are hard to diagnose.
I spent a month working around this before figuring out what was happening. Iván's work setup at the time wasn't on Google Workspace, and his Granola kept dropping calendar context, missing scheduled meeting triggers, and sometimes failing to associate transcripts with specific calls. He ended up using his personal Gmail for testing, which obviously isn't a real solution.
For solo users on Gmail this never comes up. For teams on Microsoft Outlook it's a non-starter. Otter, Fireflies, and tl;dv all treat the major email providers as equally first-class. Granola treats Google as default and everything else as a compatibility problem the user has to engineer around.
Convo doesn't carry this lock-in. It's email-provider-agnostic from day one. If your team is split between Workspace, Outlook, Zoho, or custom domains, Convo treats all of them the same way.
2. There's no way to push notes anywhere automatically
For a $14/user/month tool, this is the limit that surprised me most. There's no native way to sync a Granola note into Google Docs automatically. Action items don't pipe into your task manager without manual copying. The Notion and HubSpot integrations on the Business plan are real, but each meeting still requires you to choose where it goes.
In practice: after every client call I want shared, I copy the enhanced note out of Granola, paste it into a Google Doc, clean up the formatting, share the doc. Five minutes per meeting that I shouldn't be spending. Across a week of client calls it adds up to an hour I'm handing back to manual labor.
Most tools at this price ship some form of automatic distribution. Fireflies has a Slack pipe that just works. With Otter, the Notion sync is one-click. Granola has the parts but expects you to do the assembly after every call.
This is the kind of friction Convo was built to remove. Follow-up emails draft themselves within 30 seconds of the call ending. Action items route to the right person in Notion, Slack, or your CRM automatically. The $14.99/month covers the assembly Granola asks you to do by hand.
3. Nothing happens during the meeting itself
This is the one that mattered most to me, and probably matters most to anyone reading this who runs sales calls.
Last March, a prospect on a demo asked how we handle a competitor that drops their pricing by 30%. I had a good answer to this question. I'd worked it out six weeks earlier on a completely different call, with a different prospect, in a very similar moment. Granola had the transcript of that earlier call somewhere in its history. But Granola doesn't surface things to you during a call. That's not what it does. So I gave a worse answer than the one I already had figured out, and the deal stalled.
The fully enhanced notes from that earlier prospect call arrived in my inbox 90 seconds after the demo ended. They were great. They were also too late.
This is the gap Convo was built to close. Real-time context surfaced from past meetings. Suggested responses to objections you didn't prepare for. Help arriving while the conversation is still happening, not 90 seconds after it ends in the follow-up email. I'll admit the obvious bias here. I co-founded Convo. The structural observation is real regardless of which tool you pick. Documentation tools help you remember. They don't help you perform. Granola is excellent at remembering. It doesn't try to do the other thing.
The most important limitation in this post, condensed: Granola helps you remember meetings. Convo helps you in the moment they're happening. Bot-free capture in both. Try Convo free for 7 days.
Two opposite bets about AI in meetings
Granola's bet is that you're the one doing the cognitive work during the meeting. Your typing is the evidence. The AI's job is cleanup afterward. Convo's bet is the conversation itself is the cognitive load. The AI tries to help in the moment, not in the inbox 90 minutes later.
The practical difference looks like this:
| Granola | Convo | |
|---|---|---|
| Real-time help during the call | No | Yes |
| Bot-free capture | Yes | Yes |
| Works outside Google Workspace | Spotty | Yes |
| Cross-meeting context surfaced live | No | Yes |
| Auto-distribution (follow-up emails, action items routed) | Manual or via Zapier | Native |
| Audio processing | Cloud (training opt-in by default) | Local by default |
| Speaker attribution in 3+ person calls | Weak | Strong |
| Free tier | Basic, 25 notes total | 7-day trial |
| Starting paid price | $14/user/month | $14.99/month |
| Available on | Mac, Windows, iOS | Mac, Windows, Linux |
A privacy thing nobody else flags
Every Granola review I read in preparation for writing this one covered the same three topics: the workflow, the bot-free capture, the pricing. Almost nobody covers what comes next.
By default, Granola uses your meeting data to train its AI models. Individual users can opt out in settings. Organization-wide opt-out is gated to the Enterprise tier at $35 per user per month. On Basic and Business, every meeting contributes to training unless each person on your team flips the switch themselves.
There was also a piece in TechBuzz in 2025 about shared meeting notes being publicly accessible by default. Granola tightened that setting after the coverage. The underlying default of open before locked, training before private, is worth knowing about if you handle conversations that should stay closed.
I'm not making this bigger than it is. Most AI products train on customer data unless told otherwise. The reason I'm flagging it is that Granola positions itself as the privacy-conscious option in the category, which it mostly is, and the place where that positioning has the weakest defense is here. Convo processes audio locally on your machine by default, with no Enterprise paywall on the privacy opt-out, which is the architectural difference that matters for teams that take this seriously.
What Convo does instead
This is Convo's blog, so I'll be direct: Convo gives you the answers to most of Granola's limitations. Different design choices, different audience, but the gaps above are exactly the gaps we built Convo to close.
Convo makes the opposite bet from Granola about what AI should be doing while a meeting is happening. Granola assumes you're handling the cognitive work and the AI cleans up after. Convo assumes the conversation itself is the cognitive load and the AI should help while the call is still running.
While Granola enhances your notes after the call, Convo surfaces help during it. While Granola requires Google Workspace to work cleanly, Convo works across every email provider out of the box. While Granola asks you to copy-paste notes to share them, Convo distributes them automatically. Same bot-free capture in both products. Opposite philosophies about when AI should be doing its job.
What that looks like in practice:
Real-time suggestions during the call. When a prospect raises an objection you didn't prepare for, Convo surfaces a response pattern that's worked in similar moments. When a client references a meeting from six weeks ago, Convo pulls context from that call onto your screen within seconds. When you blank on pricing, Convo has the answer ready. The panel is visible only to you, never to other participants in the meeting.
Cross-meeting memory that surfaces live. Convo remembers every conversation you've had through it. Ask "what did Sarah commit to about the timeline last quarter?" and you get the actual quote and the link to that meeting. Granola's chat feature does some of this on past meetings after the fact, but not in the live moment when you'd actually use the answer.
Bot-free capture, same as Granola. Audio is captured directly from your machine. Nothing joins the meeting as a participant. You get the same client-facing benefit Granola pioneered, without making it the entire pitch.
Local audio processing by default. Audio doesn't get sent to a third-party cloud unless you specifically enable cloud features. No org-wide opt-out toggle to remember. Privacy isn't an Enterprise upsell, it's the default.
Works across email providers. Convo doesn't care if you're on Google Workspace, Outlook, Zoho, or a custom domain. Calendar integration is genuinely cross-platform from day one, not a Workspace-first product with workarounds bolted on.
Real auto-distribution. Follow-up emails get drafted within 30 seconds of the call ending. Action items route to the right person in Notion, Slack, or your CRM automatically. No Zapier maintenance required after every call.
The product runs on Mac, Windows, and Linux. Starter is $14.99/month. The 7-day free trial gives you full access to everything in the Professional plan.
The post you're reading is on Convo's blog for a reason. If real-time help during meetings is the thing you've actually been needing, start your free 7-day trial and see the difference inside a week.
The 25 notes you don't realise you're using
Pricing is the part I expected to be straightforward. It isn't.
Basic is free. Business is $14/user/month annually. Enterprise is $35/user/month. Normal so far.
What's not normal: Basic caps your meeting history at 25 notes total. Not 25 a month. Twenty-five, ever. You hit it and you delete history or upgrade.
| Plan | Price | Meeting history |
|---|---|---|
| Basic | Free | 25 notes total |
| Business | $14/user/month (annual) | Unlimited |
| Enterprise | $35/user/month | Unlimited + org-wide privacy controls |
The February 2026 rebrand made this worse. Granola's free plan used to be significantly more generous. The 25-note cap was added during the same rebrand that introduced the new pricing tiers and the team-customer pivot. Reporting from community forums suggests over 60% of users expressed dissatisfaction with the redesign, which is a lot of churn-pressure for a category-defining unicorn to absorb. Worth knowing if you're picking a long-term tool: Granola is in active business-model evolution, and the next pricing change is probably already in someone's roadmap doc. Convo's pricing has been stable since launch and the 7-day trial gives you full access to everything, which is a different posture from Granola's 25-note Basic plan.
For context across the field: Fireflies Pro is $10/user/month, Otter Pro is $16.99, Fathom Premium is $19, Convo Starter is $14.99. Granola's Business tier sits in the middle of the pack on price, once you've burned through the 25 free notes.
Who this is right for
Bullet list, because anything else would be performance:
- You run 1:1 client calls or small (2 to 3 person) meetings most of the time
- You're already the kind of person who types notes during conversations
- You're on Mac, Windows, or iOS (Granola still has no Android app)
- You're on Google Workspace, or willing to migrate
- Visible recording bots have changed the tone on calls you've been on
- Better post-meeting documentation is the goal, not real-time help
- You don't need to play audio back to verify what was said
For this person, Granola is genuinely the best in its category. There isn't another tool that does the write-then-enhance loop as cleanly, with bot-free capture, on desktop apps that feel native rather than packaged. (For the same person who also wants help while the call is happening, Convo is the only product I'd add to the shortlist alongside Granola.)
Who this isn't right for
- Team meetings with three or more participants where attribution matters. Otter does this better.
- You need real-time help during the call. That's Convo territory.
- You have Android users on your team. Granola keeps saying the app is coming. Until it ships, there are gaps.
- Deep Salesforce or HubSpot CRM sync is essential to your workflow. Fireflies and Avoma are stronger.
- You need audio playback for legal, healthcare, or journalism use. Any cloud-recording tool serves you better.
- You want a real privacy opt-out without paying $35/user/month. The Enterprise paywall is where the actual opt-out lives.
- You're on Microsoft Outlook or a non-Google email setup. Switching tools costs less than engineering around Granola's Workspace lock-in.
The Fathom alternatives roundup covers the wider field if Granola isn't the right shape.
Verdict
Granola is a good product built around a clear bet, and the consequences are mostly defensible if you accept the premise.
Bot-free capture makes sense once you assume the user is already typing. The Google Workspace dependency is harder to defend in 2026, but it follows from Granola being built for Mac-Google-ecosystem solo operators. The missing auto-distribution is a real cost at $14/user/month, but it's consistent with the "you do the typing, we do the cleanup" philosophy. The privacy defaults are where I'd push back hardest, and the gap between Granola's positioning and the org-wide opt-out paywall is real.
For the user Granola was built for (solo operator, Google-ecosystem-native, 1:1 calls, the kind of person who takes notes during meetings anyway), it's the cleanest tool in its category.
For the user this post is actually written for (the one who wants help during the meeting rather than better notes after), Convo exists for that exact gap. Real-time help during the call. Cross-meeting context surfaced live. Local audio processing by default. Email-provider agnostic. Auto-distribution of notes and action items without Zapier maintenance. All the things Granola structurally doesn't do, in a tool that's bot-free in the same way.
If real-time help during meetings is the gap Granola can't fill, try Convo free for 7 days. Same bot-free capture, with suggestions while the call is actually happening.
Frequently Asked Questions
Is Granola AI worth it in 2026? For solo operators and small teams who run mostly 1:1 client calls and want bot-free notes, yes. The $14 Business plan is competitive with Fireflies, Otter, and Fathom, and the workflow is meaningfully different. For team meetings with 3+ people, deep CRM workflows, or real-time help during calls, look at Convo, Fireflies, or Avoma instead.
Is Granola AI accurate? Transcription accuracy lands at 90-95% in clean audio with one speaker at a time. Multi-speaker accuracy drops noticeably in calls with 3 or more participants. Speaker identification works in 1:1s and small calls, breaks down in larger meetings. Otter still leads on raw accuracy and attribution.
How does Granola compare to Otter? Granola wins on bot-free capture and the note-writing workflow. Otter wins on transcription accuracy, search across past meetings, and speaker identification. If you care about archives and reference, Otter. If you care about how clients perceive recording, Granola. See Granola vs Otter vs Fathom for a full breakdown.
Does Granola record audio? No. Granola only captures and stores the transcript. There is no audio or video recording you can play back. For some users this is a privacy benefit. For others, the inability to verify a transcript against the original audio is a real workflow gap.
Does Granola work on Windows? Yes. As of May 2026, Granola ships native apps for Mac and Windows. The Windows app launched after the Mac app and has caught up over the last year. Mobile is iOS only. There is no Android app and no public release date.
Is Granola safe to use for sensitive conversations? Granola captures audio locally and doesn't put a visible bot in your meeting, both of which are privacy wins. The two caveats: meeting data is used to train Granola's AI models by default (individual opt-out in settings, organization-wide opt-out only on the $35 Enterprise plan), and the company has had documented issues with shared notes being more public than users expected. For healthcare, legal, or financial conversations where these matter, evaluate carefully.
What's the cheapest paid Granola plan? Business at $14/user/month on annual billing. There is no monthly-only option below that. The Basic plan is genuinely free but caps your meeting history at 25 notes total, which most active users hit inside a month.
Are there alternatives that help during the meeting, not just after? Yes. Convo is the main one. It also runs bot-free on Mac, Windows, and Linux, but surfaces real-time suggestions during the call rather than only documenting afterward. Granola opened the bot-free category. Convo extended it to real-time help. The Fathom alternatives roundup covers the wider field.
Should I use Granola if I'm on Microsoft Outlook? Probably not. Granola is built around Google Workspace and the integration on non-Google email setups (Outlook, Zoho, custom domains) is brittle. Calendar context drops, meeting triggers get missed, and transcripts sometimes fail to associate with specific calls. Otter, Fireflies, and Convo all treat email providers as equally first-class. If your team is on Outlook, the cost of switching tools is lower than the cost of working around Granola's Workspace lock-in.
Granola raised $125M in 2026. Does that affect users? It already has. The Series C in March 2026 came alongside a strategic pivot from prosumer notetaker to enterprise context layer, the February 2026 rebrand that added the 25-note cap on Basic, the launch of Spaces (team workspaces with access controls), and the MCP integration that pipes meeting data into Claude and ChatGPT. The roadmap is moving toward enterprise teams and away from solo users. If you're a solo user on the free plan, expect the deal to keep changing in ways that favour upgrade.
Does Granola work for teams or just solo users? Honest answer: better for solo users today, despite Granola's enterprise pivot. Speaker attribution breaks at three or more participants, there's no native auto-distribution of notes to shared destinations, and the Google Workspace dependency creates friction for any team mixing email providers. Spaces (the team workspaces feature) exists but it sits on top of these structural limitations. For teams where attribution and distribution matter, Otter or Fireflies is currently a better fit.
Learn more about this topic with AI
Ready to transform your meetings?
Join professionals using Convo to feel confident in every conversation.
Download for MacCONTINUE READING