Amidst the booming wave of AI applications, Granola stands out with its 70% user retention rate and valuation of $250 million, making it one of the most successful AI note-taking apps in Silicon Valley. This article will reveal the secret of Granola’s success and the product development cognition behind it.
In the AI industry, AI note-taking applications are not uncommon. Before the release of ChatGPT, there were already many products. But even so, there is still an AI note that has come out.
It’s Granola.
This product, which has only been launched for one year, has been the most successful AI note-taking application in Silicon Valley and has won the favor of many tech tycoons.
Official statistics show that more than half of Granola users hold leadership positions, including founders and executives of unicorns such as Vercel, Ramp, Roblox, and others. From the data point of view, the number of users of this product has increased by 10% every week, and the monthly retention rate is as high as 70%.
On May 14, Granola announced the completion of a $43 million Series B funding round valued at $250 million. This valuation alone is quite capable in the field of AI notes.
So, what exactly is different about Granola AI?
01 Give users control, AI plays the “second brain”
Before Granola, AI note-taking products generally had two ways, either they would let AI take over the meeting minutes entirely, or AI would directly generate relevant meeting minutes.
Unlike these AI notes, Granola’s core principle is that AI should enhance human capabilities, not replace them.
In the view of Granola founder Chris Pedregal, the difference between the two is that the former simply outsources daily tasks to artificial intelligence while retaining human judgment and creativity.
Granola gives users enough control when it comes to the scene.
Granola’s scenario is that in a meeting, users can take any notes or flashes of ideas at any time, and Granola captures and transcribes the conversation in real time.
After the meeting, it will further refine and enrich the transcription based on your recorded notes, making the final note more complete and valuable. In this way, users don’t need to record everything in the meeting in detail, but can focus on those key insights or your unique thinking and judgment, and leave those tedious recording tasks to AI.
What does a product manager need to do?
In the process of a product from scratch, it is not easy to do a good job in the role of product manager, in addition to the well-known writing requirements, writing requirements, writing requirements, there are many things to do. The product manager is not what you think, but will only ask you for trouble, make a request:
View details >
For example, in a meeting, you write down a number, and Granola extracts from the conversation in real time and adds context about the meaning of that number, making the final note more complete and valuable.
The benefits of this approach are twofold:
First, it maximizes the understanding of user context and priorities than most summarization tools, while capturing user intent directly from the source by recording low-fidelity, keyword-based notes during meetings, rather than enforcing fully automated or generic output.
Second, compared with other AI notes, Granola is closer to the real scene of user meetings.
For example, most users will not write detailed and complete notes during the meeting, but will only hastily write down a few scattered keywords and tell themselves to sort them out later.
Granola has also tried to write complete notes in real time. But the team found that no matter how well the notes were written, the real-time generation was compelling to watch, and the results were very distracting. This is a departure from Granola’s original intention. Granola was originally intended to make you more focused on the meeting, but it backfired, with people staring at their notes and changing them if they weren’t satisfied.
Granola cleverly recognizes the clutter and imperfections of real note-taking and offers a seamless approach to improvement. As one user explained why Granola resonates:
“It didn’t ask me to change the way I worked. It embraces my messy, imperfect note-taking habit and makes it better. This simplicity resonated with me because it felt like it was designed just for me. ”
Behind the user-led control is Granola’s unique product positioning.
In the words of the founder of Granola AI, Granola is not just a recording tool, but an intelligent assistant that helps us understand, connect, and utilize information. This is somewhat similar to the concept of a “second brain”.
It’s like giving you memory and connecting an external hard drive. The physical limitations of our brain determine its memory capacity, and these tools actually provide you with additional storage space.
Chris Pedregal believes that all those who develop AI tools must consciously work in this direction. Because users just want to outsource all repetitive, boring, mindless work, but don’t want to outsource judgment.
02 From AI note-taking to team collaboration platforms, what did Granola do right?
In addition to the product concept, Granola’s business path evolution is also interesting. Recently, Granola launched version 2.0, positioning itself from a personal note-taking tool to a team collaboration platform.
This shift stems from Granola’s choice of meeting scenarios. There are two values in choosing a meeting as an entry scenario:
Firstly, AI capabilities are highly matched to the needs of the meeting scenario. Among the many key advantages of LLMs—code generation, search, and summarization—this ability to distill long-form notes greatly streamlines the meeting process and resonates with users.
Second, meetings are a natural way to engage users and are easier to integrate into user workflows. For example, the Granola team found that the most valuable information in the company is not in static documents or Wikipedia, but in the day-to-day conversations between teams.
In order to make the product more suitable for the needs of the conference scene, Granola has also made a lot of design in terms of security.
For example, the Granola team decided not to record or store any audio data. Although Granola listens to the audio in real time and transcribes it, the original audio file itself is not saved.
This means that Granola far outperforms AI products that will join your meetings and record and store audio and video.
That’s why Granola is expanding from a simple meeting note-taking tool to a team collaboration platform. Specifically, it seeks to change the way teams acquire, share, and leverage collective knowledge with new capabilities:
Shared Team Folders: Create dedicated spaces for sales calls, customer feedback, hiring loops, weekly syncs, and more that anyone on your team can access (even without a Granola account);
Chat with Folders: Query entire meeting minutes folders using top-notch reasoning models, with AI providing insights that reference specific meetings and notes as sources;
Enterprise Collaboration: Business and enterprise users can browse any public folder within their domain – perfect for competitive intelligence, customer success, or new employee onboarding;
Slack Integration: With a one-time connection to Slack, Granola keeps your entire team on top of the meeting and publishes a concise summary and a “Chat with this meeting” button on your chosen channel at the end of the call.
Behind this transformation, it also reveals the Granola team’s judgment on AI product competition. In a previous interview, Chris Pedregal mentioned:
Future AI products can be roughly divided into two categories, one is products that perform low-frequency cutting tasks that do not require too good tasks, and most consumption scenarios may belong to this category, and this part of the demand will be eaten up by general-purpose assistants like Claude. The other category is high-frequency scenarios that require excellent performance, which is the domain of professional AI tools. Granola belongs here.
03 4 perceptions behind Granola’s success
Behind Granola’s success, Chris Pedregal also has a deep understanding of AI product development. In a previous interview, Chris shared his four unspoken rules for creating successful AI products.
First, don’t waste time on issues that are about to disappear.
Typically, the problems of an AI product arise in two ways: one is the problem that will be automatically solved when the next model is released, and the other is the challenges that will still exist no matter how intelligent the model becomes.
For application layer entrepreneurs, it’s easy to fall into the trap of “misinterpreting the problem”: you work hard to solve an immediate problem, and as soon as the next version of GPT comes out, the problem will no longer exist. So, don’t waste time on issues that are about to disappear.
For example, the first edition of Granola could not record meetings longer than 30 minutes. The strongest model at the time was OpenAI’s DaVinci, with a context window of only 4,000 tokens.
According to the logic of traditional products, the team should immediately prioritize solving this problem, but Granola did not do this and focused more on improving the quality of the notes. This decision turned out to be the right one, and the quality of notes is one of the reasons why users love Granola the most.
Second, marginal cost is an opportunity.
In the past, the marginal cost of Internet products was extremely low, and good products could easily achieve user scale growth.
But this law does not apply in the field of AI. Due to the existence of model costs, each additional user brings an increase in marginal costs. This means that even large manufacturers cannot blindly expand users.
This is an opportunity for entrepreneurs. While you can’t compete directly with larger companies, you can choose to provide a better product experience for a small number of users. This is more feasible than serving tens of millions of users.
Third, context is king.
Instead of thinking of AI models as “tools for executing commands,” think of them as interns on their first day at work. An intern with independent thinking ability, rather than teaching him what to do, is more important to give him enough to understand your “context” and let him truly understand the intention. In the era of AI, “context window selection” will become the core concept.
Fourth, solve the problem of specific scenarios.
Universal tools like Claude and ChatGPT are good enough now. How to compete with them? The only way is to make it “narrow” enough. Choose a very specific scene and do it to the extreme in this scene.
When you polish in the process of landing specific scenes, there will be a lot of Know How. For example, whether the meeting reminder is smooth, whether the echo elimination is good enough, and the length of the meeting minutes are appropriate, these need to be optimized through continuous user feedback, which cannot be solved by the model.
