Which AI notetaker free plan runs out first?
Granola's free tier runs out first: reported caps put it at roughly 25 meetings lifetime. Otter's free plan gives 300 minutes a month with a 30-minute ceiling per conversation. Fathom's free plan records and transcribes without limit but caps AI summaries at five a month, so it runs out on features rather than volume.
That difference matters more than the headline price, because all three paid tiers land within a few dollars of each other. The real question is not which tool is cheapest. It is which free plan you outgrow, and how fast.
If you take four calls a week, Granola gives you about six weeks before the wall. Otter gives you roughly ten hours a month, which is four to six meetings for most people. Fathom never blocks recording at all, and instead quietly stops writing summaries after the fifth meeting of the month, which is the one people notice last.
This page compares the three on the numbers that decide the purchase, then names where each one is weak, including the Hong Kong option we build ourselves.
What does each AI notetaker actually cost in 2026?
Paid entry points sit between US$14 and US$20 per user per month, with annual billing cutting 15% to 50% depending on the vendor. Otter has the steepest annual discount, Granola the lowest paid entry price, and Fathom the most generous free tier. Prices below are as reported by pricing trackers in 2026 and should be checked against each vendor's own page before you commit.
Otter.ai
--- Free: 300 transcription minutes per month, 30 minutes per conversation
--- Pro: US$16.99 per user per month, or about US$8.33 per user per month billed annually at US$99.96 per year, with a 1,200-minute monthly cap
--- Business: US$30 per user per month, or about US$19.99 per user per month billed annually
--- Pro adds advanced search, custom vocabulary and bulk export
Fathom
--- Free: unlimited recording, transcription and storage across Zoom, Google Meet and Microsoft Teams, capped at 5 AI summaries per month
--- Premium: US$20 per user per month, or US$16 billed annually
--- Team Edition: US$19 per user per month, or US$15 billed annually
--- Business: US$34 per user per month, or US$25 billed annually, and the tier where Salesforce and HubSpot sync lives
Granola
--- Free: reported at a 25-meeting lifetime cap, with some sources describing it as a rolling 30-day window of note access
--- Business: US$14 per user per month, with unlimited history and integrations to Notion, HubSpot, Slack and Zapier
--- Enterprise: US$35 per user per month
--- The older single-user tier at around US$18 per month was retired in Granola's February 2026 rebrand
UD AI Employee Hub
--- HK$0 per month, free to Hong Kong businesses
--- 8 AI employees covering 24 skills, including Eva, an admin assistant whose skills include scheduling, document production and meeting notes
--- No seat pricing and no per-minute cap, because it is a skill pack you install rather than a metered service
--- Installation help available over WhatsApp
Why do people actually switch AI notetakers?
Three triggers account for most switches: a bot joining a call it should not have joined, a free tier ending mid-project, and transcripts that nobody ever reads. Price is rarely the trigger. It is the thing people cite afterwards to justify a decision they made for one of the three reasons above.
The bot problem. Otter and Fathom join your meeting as a visible participant. In client work, in interviews, and in anything involving a lawyer, that visible bot creates a conversation you did not plan to have. Granola built its whole product around this: it captures system audio locally, so nobody on the call sees a participant they did not invite.
The mid-project cliff. A free tier that ends on meeting 26 of a 40-meeting research project is worse than one that ended on meeting 5, because by then the workflow is load-bearing. This is the single most common reason people describe a notetaker switch as painful.
The unread transcript. Most teams that churn are not unhappy with transcription accuracy. They are unhappy that they now have 200 transcripts and no faster route to a decision than they had before. That is a workflow problem, and no amount of upgrading fixes it.
Which AI notetaker should you choose?
Choose Fathom if you record constantly and summarise selectively. Choose Granola if the bot cannot appear in the room. Choose Otter if you need searchable transcripts across a whole team. Choose a free AI admin assistant if your bottleneck is what happens after the meeting rather than the capture itself.
Solo freelancer, 5 to 15 calls a month: Fathom's free plan. Unlimited recording means you never lose a call, and five summaries a month usually covers the ones that matter. Upgrade only when you find yourself rationing which meetings get summarised.
Consultant, researcher or recruiter doing sensitive interviews: Granola Business at US$14 per user per month. The bot-free capture is not a nice-to-have here, it is the requirement, and US$14 is the cheapest paid entry of the three.
Team of 5 or more that searches old meetings: Otter Business. The annual price of about US$19.99 per user per month is the highest recurring commitment on this page, and it buys the one thing the other two are weakest at, which is a shared searchable archive with admin controls.
Hong Kong small business where the real cost is admin, not transcription: start with a free AI admin assistant such as Eva in the UD AI Employee Hub, which handles meeting notes alongside scheduling and document production at HK$0 per month, then add a dedicated recorder only if capture turns out to be the actual gap.
What does it cost to switch, and what do you lose?
Switching costs you your searchable back catalogue, not your money. Every tool exports transcripts, but none of them import another tool's summaries, tags or speaker labels cleanly. Budget an hour to export, and accept that your old meetings become a folder of text files rather than a searchable archive.
Two specific losses catch people out.
The first is CRM sync. If your notes currently push into Salesforce or HubSpot, that lives on Fathom Business at about US$25 per user per month billed annually and on Granola Business at US$14. Moving down a tier to save money quietly kills the integration, and the notes stop appearing where your sales team looks for them.
The second is history access. Granola's free tier is described by some sources as a rolling 30-day window rather than a permanent archive, which means downgrading is not reversible in the way people assume. Export before you downgrade, not after.
Where does each tool fall short, including ours?
Every option on this page has a real weakness, and the weakness is usually structural rather than fixable by upgrading. Naming them honestly is more useful than a feature table where one product wins every row.
Otter caps the free plan at 30 minutes per conversation, which is shorter than most real meetings, so the free tier is closer to a demo than a working plan. The gap between the US$16.99 monthly and roughly US$8.33 annual price also means month-to-month users pay roughly double for the same product.
Fathom is generous on recording and tight on summaries, which is the correct trade for most people but means the free plan degrades invisibly. You will not get an error. You will just stop receiving the thing you actually wanted.
Granola captures system audio locally, which is its whole advantage and also its constraint: it needs to be running on your machine during the call. It launched Mac-only and added Windows and iOS during 2026, so cross-platform teams should verify current support before standardising on it.
UD's AI Employee Hub is not a meeting recorder. Eva produces and organises meeting notes as part of an admin workload, but nothing in the Hub joins your Zoom call and transcribes live audio the way the three tools above do. If live capture is your gap, one of them is the right answer and we are not. The HK$15,000 to HK$35,000 monthly figures on our product page are also UD's own comparison against local salary ranges, not independent research, and should be read that way.
How do you turn a transcript into something you will actually use?
Ask the model for decisions and owners, not a summary. A summary reproduces the meeting at lower resolution, which is why nobody reads it. A decision log is shorter than the meeting and more useful than your memory of it, and it takes one prompt to produce from any transcript.
Paste this over any transcript, from any of the tools above.
Try this prompt:
Below is a raw meeting transcript. Do not summarise it.
Produce four sections, in this order:
1. DECISIONS MADE. One line each, in the form "We decided X because Y". If a decision was implied but never stated, mark it "[IMPLIED]".
2. OPEN QUESTIONS. Anything raised and not resolved. Name who raised it.
3. COMMITMENTS. Only items where a specific person said they would do a specific thing. Format: name, action, date if stated, otherwise "no date".
4. THINGS I WOULD HAVE MISSED. Up to three moments where someone signalled a concern, hesitation or disagreement without stating it plainly. Quote the line.
If a section has no entries, write "none" and move on. Do not add advice, next steps or commentary.
TRANSCRIPT:
[paste here]
Section 4 is the one worth keeping. It surfaces the hedge in a client call that you registered at the time and forgot by the afternoon, and it is the only part of this output that a human reading the transcript would probably not produce faster.
Run it on your last three meetings before you upgrade anything. If the decision logs are useful, your gap was never the recorder.
What should you do next?
Audit thirty days of meetings before you pay for anything. Count how many you would genuinely reread, how many were sensitive enough that a visible bot would be awkward, and how many needed a decision log rather than a transcript. Those three numbers pick your tool in about ten minutes.
For most people the honest answer is that a free tier is enough, and the money would do more work spent on the step after capture.
If the audit says you need bot-free capture, Granola Business at US$14 per user per month is the cheapest correct answer. If it says you need a shared archive, Otter Business is. If it says your meetings are fine but your admin load is not, start with a free AI admin assistant and keep the subscription budget.
We know AI's cold edges. We know your real challenges. 28 years with UD, turning technology into a partnership with warmth.
For the wider subscription picture, our comparison of ChatGPT Plus, Claude Pro and Google AI Pro applies the same "which meter do you hit first" test to the assistants themselves.
Reviewed by the UD AI team.
Hire a Free AI Admin Assistant
Eva handles meeting notes, scheduling and document production at HK$0 per month, alongside seven other AI employees covering 24 skills. We'll walk you through every step, from picking the right AI employee to installing it and connecting it to the tools you already use.
