You can build a fully automated AI meeting summary workflow in under two hours using tools like Otter.ai, Make.com, and Notion — and once it’s running, you’ll never manually write up a meeting recap again. This guide walks you through every step, including the parts that actually trip people up.
Why Meeting Summaries Are the Perfect First Automation
Most teams lose somewhere between 15 and 30 minutes after every meeting just reconstructing what was said, who owns what, and what the decisions were. Multiply that across a 10-person company with five meetings a week and you’re hemorrhaging real time. The good news: this is one of the cleanest automation wins available right now because the inputs are predictable (audio or transcript), the output is well-defined (summary + action items), and the stakes of a small error are low. Nobody’s going to sue you if a bullet point is slightly off.
Before you build anything, it helps to understand how AI actually retrieves and processes context from longer inputs — if that’s new territory, the explainer on how AI finds the right answer using RAG is worth five minutes of your time.
What You’ll Need Before You Start
Here’s the stack I’d recommend for a US-based team that uses Google Meet or Zoom:
- Otter.ai — transcription and initial summary generation
- Make.com — the automation glue that connects everything
- Notion (or Slack, if you just want summaries pushed to a channel) — where the final output lives
- A Google account if you’re using Google Calendar for meeting triggers
You don’t need a developer. You need about 90 minutes and a willingness to click around.
Step 1: Set Up Otter.ai to Automatically Join and Transcribe Meetings
Otter.ai’s Business plan ($20/user/month as of mid-2025) includes OtterPilot, which automatically joins your Google Meet or Zoom calls, transcribes them in real time, and generates a summary when the meeting ends. The free tier exists but tops out at 300 minutes per month total, which disappears fast if you have more than a handful of meetings weekly. I’d skip the free tier entirely for this use case — you’ll hit the cap, the workflow will break, and you’ll wonder why.
How to Configure OtterPilot
- Go to otter.ai → Settings → OtterPilot
- Connect your Google Calendar or Outlook Calendar
- Toggle on “Automatically join meetings I’m invited to” — be thoughtful here, because it will join every calendar event with a video link, including sensitive ones
- Set the summary format: I recommend “Action Items + Key Points” over the default paragraph summary
One real limitation worth knowing: Otter.ai struggles with heavy crosstalk and strong accents. If your team has non-native English speakers or you run panel-style discussions, the transcript accuracy drops noticeably and you’ll want to budget time for light editing. It’s not a dealbreaker — just set expectations correctly.
Step 2: Get Otter.ai to Export Summaries Automatically
After a meeting ends, Otter.ai sends a summary email to all participants. That’s fine but not enough — you want the data in a structured place, not buried in inboxes. Here are two ways to extract it:
Option A: Use Otter.ai’s Native Integrations
Otter.ai has direct integrations with Notion, Salesforce, and HubSpot under their Business plan. If you just want summaries to land in Notion automatically, go to Settings → Integrations → Notion, authenticate, and pick the database where notes should go. Done. This takes about four minutes and requires zero Make.com setup. The downside: you get limited control over formatting and you can’t easily branch the logic (like sending sales call summaries to HubSpot but internal standups to Notion).
Option B: Route Through Make.com for Full Control
This is what I’d recommend for any team that wants conditional logic. Make.com’s free tier gives you 1,000 operations per month, which is enough for testing. Their Core plan is $9/month for 10,000 operations, which covers most small teams easily.
The trigger in Make.com: use the Email module set to watch your inbox for emails from no-reply@otter.ai with the subject line containing “Your Otter Summary.” Parse the email body using a text parser module to extract the summary and action items. It sounds fiddly, but Otter’s summary emails are consistently formatted — the parser works reliably once you set it up. If you’ve never built a scenario in Make, the tutorial on building a lead scoring automation in Make.com is a solid primer on how the interface works.
Step 3: Structure the Output in Notion
Raw summaries dumped as plain text get ignored. Structure matters. In Make.com, after you parse the Otter email, add a Notion → Create a Page module. Map it like this:
- Title: Meeting name + date (pull both from the email subject line)
- Summary body: The parsed summary text
- Action Items: A separate Notion property, either a multi-line text block or a checkbox list if you parse them individually
- Attendees: Also in the Otter email — map this to a Notion “People” or plain text property
- Tags: You can add conditional logic in Make to tag notes as “Sales,” “Internal,” or “Client” based on keywords in the meeting title
Notion’s free plan works fine for this. The $10/user/month Plus plan adds version history, which is genuinely useful if you want to track how action items evolve over time.
Step 4: Add a Slack Notification (Optional but Recommended)
Even if Notion is your system of record, most teams actually live in Slack. Add a Slack → Post Message module at the end of your Make.com scenario. Format the message as a short digest: meeting name, three-line summary, and a link to the full Notion page. Keep it under 200 words. People will actually read it.
Set the Slack destination channel based on that same keyword logic — sales calls go to #sales, client meetings go to #account-management. This is the part that impresses people who see it for the first time. It looks like someone manually curated it. They didn’t.
Step 5: Handle Edge Cases So the Workflow Doesn’t Break
Every automation has failure modes. Here are the three I’ve seen most often with this specific setup:
Otter Didn’t Join the Meeting
This happens when the calendar event was created less than 10 minutes before the meeting started, or when the video link format isn’t recognized. Fix: add a standing 15-minute buffer when creating last-minute calls, and always use the full Zoom or Meet URL in the calendar invite — not a shortened link.
Make.com Scenario Fails Because the Email Format Changed
Otter occasionally updates their email templates. Add a Mailhook or email logging step so you can see raw incoming emails if a scenario suddenly stops firing. Takes 10 minutes to set up and saves hours of debugging later.
Notion Rate Limiting
Notion’s API allows roughly three requests per second. If your company runs five simultaneous meetings that all end at the hour mark, you might hit this. Add a small Sleep module (5-10 seconds) between the parse step and the Notion creation step as a buffer. Not elegant, but it works.
Real Results to Expect
A three-person product team I know ran this workflow for 30 days and tracked the time savings manually. They averaged 4.2 meetings per day across the team. Before the automation, each meeting generated about 12 minutes of post-meeting note cleanup. After: under two minutes to skim the Notion page and confirm accuracy. That’s roughly 42 minutes saved per day across the team — not life-changing in isolation, but the quality of their Notion meeting archive improved dramatically because every meeting was captured, not just the ones someone remembered to write up.
If you’re curious whether this kind of automation is worth the investment or you’ve heard conflicting things about what AI can realistically do for small teams, it’s worth reading through the breakdown of business AI myths that cost companies money — some of the skepticism is warranted, but not all of it.
My Honest Recommendation
Start with Option A — Otter.ai’s native Notion integration — and run it for two weeks before deciding if you need Make.com at all. Most teams with straightforward needs will find the native integration sufficient and won’t need to touch an automation platform. Only add Make.com if you need conditional routing, CRM sync, or Slack formatting you can’t get natively. The temptation to over-engineer this is real, and I’ve seen people spend three days building a complicated multi-branch Make scenario when a single Otter → Notion connection would have handled 95% of their use case.
If the meeting summary workflow goes smoothly and you catch the automation bug, the same logic applies to other content pipelines — the guide on automating a social media content pipeline with AI uses a lot of the same building blocks.
FAQ
Does this workflow work with Microsoft Teams?
Otter.ai supports Teams transcription, but the OtterPilot auto-join feature is less reliable with Teams than with Zoom or Google Meet as of mid-2025. An alternative worth checking out is Microsoft Copilot, which is built natively into Teams and handles summaries well if your organization is already on Microsoft 365. The Make.com and Notion pieces of this workflow still apply — you’d just swap the transcription source.
What if I don’t want Otter.ai joining sensitive meetings?
Smart concern. In OtterPilot settings, you can exclude specific calendars entirely or mark individual events as “private” in Google Calendar — Otter respects the private flag and won’t join those calls. You can also set up a keyword blocklist in the calendar event title, though that requires a small Zapier or Make workaround. Just don’t rely on the default “join everything” setting without thinking it through first.
How much does this entire setup cost per month?
For a small team of five people: Otter.ai Business is $100/month (5 × $20), Make.com Core is $9/month, and Notion Plus is $50/month (5 × $10) — or $0 if you’re already on Notion’s free tier. So roughly $109 to $159/month all-in. If your team was paying a part-time admin $18/hour to handle meeting notes for even six hours a week, the math on automation is obvious.
This article was produced with the assistance of AI, and its featured image was AI-generated. We review for accuracy, but please verify critical details.
