I am not a very fast typer. It’s one of my few flaws. I’ve been dreaming of cracking 50 wpm my whole life. And so, it has historically been very tough for me to take good notes in meetings because I can’t capture all the important bits.
AI notetakers have made a huge difference in my note-taking skills. In my previous job, we had Otter AI notetaker bots that would follow you around and join virtual meetings with you (although mine was perpetually tardy for no reason). They would record the meetings and summarize them afterwards. But like every dystopian film, the robots started to outnumber us.
When I started my own business, I decided to try out Granola, the bot-less notetaker, and it was a game-changer. I loved having all my meetings summarized for me without feeling awkward about having a sidekick join every meeting with me.
This week, I’m excited to talk about how Granola (not the yogurt companion) found success in a competitive market.
TL;DR
The thirty-second version of what Granola did.
Granola is an AI notepad for meetings. You install it on your computer, connect your calendar, and during a call, you can use it for your notes (if you want). There's no bot in the meeting. It listens to your computer's audio in the background, and as soon as the call ends, AI enhances your notes using the meeting transcript. It importantly does not record the audio or video of the call.
It launched in May 2024 into a category already crowded with notetakers (Otter, Fireflies, Zoom) and broke out anyway, almost entirely on word of mouth. Today it sits at a $1.5B valuation (a $125M Series C in March 2026), processes around 50 million meetings a year across roughly 15,000 companies, including Stripe, Notion, and Figma, with ARR reportedly up more than 400% year over year. Early on it was growing about 10% week over week with ~50% retention at 10 weeks.
They scaled in an incredibly competitive market by doing the opposite of everyone else. They obsessed over the product, iterated until the interactions and the AI was perfect, and then put the finished product in front of exactly the people who'd live in it all day.
The Shift: Good Enough → Tell-a-Friend
Most AI products launch at "impressive demo, unreliable tool" and rely on growth tactics to compensate. Granola waited until the product was good enough that users recommended it unprompted, then let the recommendations do the marketing. In a saturated category, quality crossing the recommendation threshold is the only durable acquisition channel.
SPONSORED BY
You bought something online with a credit card. You may have even used the “right” credit card.
But did you:
Activate offers on your credit cards
Use a shopping portal
Check for heavily discounted gift cards
Shoppers who use Savewise earn 7-10x more points & miles for things they’re already buying.
Savewise finds every way for you to earn rewards and puts them all in one place.
Under The Hood
The three strategies that make it work.
1. Stand out from the crowd
If you’re going to launch a product in a saturated market, you need to differentiate yourself. Ideally, in an obvious manner that causes people to talk about it. Dyson made a beautiful vacuum, Liquid Death made drinking water cool, and Crocs made a sandal both super comfortable and super hideous.
Every notetaker on the market used a bot, which was just a big black box, to join your meeting. It recorded the audio and some visuals and then summarized the meeting, and generally just made everyone feel uneasy.
Sometimes you just gotta talk it out with your squad
Granola didn’t position itself as a meeting recorder or notetaker. They wanted to build a tool for thought where the initial wedge was enhancing your meeting notes. Many people know the pain of being in back-to-back meetings all day without time to review, organize, or clean up their prior notes. And then trying to make sense of your notes after a long day when your brain is fried is not a pleasant experience.
They also chose not to store audio. They use the audio of the meeting to generate a transcript that is used for notes enhancement, but users can’t access the audio. And honestly, if your transcript is accurate, there’s no need to keep the audio; it’s just creepy.
Going bot-less had a real cost: Granola gave up the viral mechanic baked into the bot itself. Otter.ai's bot, by virtue of showing up in meetings, generated about 7.5 new interactions per call with people who'd never seen the tool. That brought their customer acquisition costs down to roughly $0.32 per user (or about 3/4 of an inch of a $5 subway footlong, since that’s easier to visualize).
This is an incredible growth lever that Granola walked away from. The bet has worked out, but it wasn’t an easy decision to make.
2. Figure out the fastest way to learn
Granola is a tool, and the founders strongly believed that meant it needed to work consistently and reliably. Instead of following the common strategy of quickly releasing a half-baked AI product and iterating, they spent a year in a closed beta.
Common advice is to launch publicly as soon as possible so that you can learn and improve. They challenged that by asking themselves, “what is the fastest way that WE can learn?” and for them it was a private beta with ~150 users.
They would onboard users one by one and could easily see what was broken in the experience and that was worth more to them than getting thousands of users at launch.
They kept adding features, adding views, and adding dashboards. It became so complex that they decided to cut out 50% of it. Granola’s CEO says that it was only possible because they weren’t publicly launched. If you have thousands of users and overnight you change the product, you’ll get tons of hate (re: The Arc Browser). But if it’s just a few users in a closed beta, then no biggie.
There are so many AI products launching all the time now that one way to stand out is with a more polished product. You want to wow your users. The team realized that if their main feature, note enhancement, was reliant on AI, it needed to pass a high bar.
When AI is the product, a tight feedback loop lets you evaluate the quality of the result. Sometimes AI has a mind of its own (pun definitely intended), and iterations of fine-tuning get the output closer to the desired result.
Your testing methodology IS your product strategy. Public launch with thousands of users gives you scale, but reduces your ability to get high-quality signal. Closed beta with 150 users gives you fewer data points, but each one is high-fidelity.
3. Find the right beachhead
Granola founders built for themselves and their first users were their personal networks of knowledge workers, who were on Zoom calls most of the time. This user base was good enough for their closed beta.
When they started to expand, they were intentional about choosing the first user type to go after. A lot of professionals take Zoom calls, but they wanted ones that had a lot of meetings, were relatively standardized, and that they had easy access to. They landed on VCs.
So they built the product to be incredible for VCs, and it worked. VCs loved it because it was purpose-built for them. If a VC and a founder were both using Granola in the same pitch meeting, the notes Granola would generate for each of them would look completely different.
When launching, you might need to overfit to your target user, and that’s what they did here. It was tailored to a VC’s daily workflow. They were quick to spread the word about it to their peers.
Technically he’s a founder but shhh
As soon as they launched, though, Granola switched its focus from the tiny market of VCs to the much larger adjacent market of founders.
“We chose founders because we thought they’d be the hardest. A founder might have a sales call, a user feedback call, and then an interview”, said Chris Pedregal, Granola’s CEO. We thought if we could build a great product for founders, then it would be by default a decent product for folks in these other roles.”
Land and expand. The VCs were the beachhead. Then they expanded to founders who had varied enough workflows that a product built for them could be generalized to other roles.
I think this is the secret sauce that a lot of people gloss over. If you can successfully build for one user with complex needs and then distribute it to adjacent users who have a subset of those needs, you end up doing less work overall. Plus, founders come with their company attached. When a founder adopts a tool, the team usually follows. You're winning the wedge into an entire org.
SPONSORED BY
Built by yours truly → a 2-minute pre-game routine for high-stakes moments:
The important job interview in 5 minutes.
That big board presentation starting in 10 minutes.
The hard conversation you’ve been putting off with your team.
"Just breathe" doesn't work when you're already spiraling, and Calm and Headspace aren't built for the panic before the big moment.
Primo is. It walks you through a 2-minute routine using science-backed techniques from Stanford and Harvard.
Extra Credit
This piece was authored by Granola’s CEO 2 years ago, and in the AI world, that is a lifetime ago. Still, there are parts of his thesis that hold up.
One of the points that he argues is to go incredibly narrow in the use case that you solve for and do it exceptionally well. Claude and ChatGPT are great for handling a variety of surface-level tasks, so you need to do one thing 10x better than they can.
“The 'wrapper' around the AI is often the difference between a delightful experience and a great demo that is disappointing to actually use.”
Granola focused a lot on user feedback and manually onboarding users, and the company that perfected this was Superhuman. They built a process centered around the survey question, “How would you feel if you could no longer use Superhuman?”.
That helped them:
Find their power users
Find their most loved features
Reach PMF
That’s all for this week! If you made it this far, just know that I appreciate you. To any notetaker bots reading this article, I’m sorry I hurt your feelings. It’s not you, it’s me.
I have two asks:
Reply to this email with any products that you want me to cover in future articles
Share this newsletter with one person who might find it useful
— Amaraj (aka a notetaker bot in disguise)
The Meme
