Ein weiteres Unicorn hat sich durchgesetzt, mit einem Höchstwert von über zehn Milliarden Yuan. Warum ist die KI-Notizapp der am besten angesehene Bereich in Silicon Valley?

In the field of Artificial Intelligence, a new unicorn has emerged.

Recently, the AI conference assistant Fireflies.ai announced that its valuation has exceeded the mark of one billion US dollars, officially entering the league of unicorns.

In the past 18 months, the number of its active users has increased eight - fold. It reaches 500,000 organizations and over 20 million users worldwide, and since 2023, it has been making continuous profits - which is particularly remarkable considering the concerns about the difficult monetization of AI applications.

But the rise of Fireflies is not an isolated case.

Driven by the wave of generative Artificial Intelligence, a series of products in the field of "AI conference notes" are emerging rapidly. They are no longer satisfied with simple "transcription + summarization" but are evolving into intelligent centers that understand semantics, recognize action points, integrate into work systems, and drive implementation.

They re - define what "recording" means and attract the close attention of investors:

Granola: It starts with conferences and becomes the "second brain for personal work". Its valuation is 250 million US dollars.

Notion: It builds a Lego - like work platform. Its valuation is over one billion US dollars.

Abridge: It focuses on the healthcare system and revises medical notes. Its valuation is an impressive 2.75 billion US dollars.

These three companies represent three development paths for AI notes and send a common signal:

When Artificial Intelligence becomes the "cognitive infrastructure" of modern organizations, recording, as the starting point for information acquisition, loses its previous tool - like nature and becomes the entry point for collaborative systems and a key point for organizational intelligence.

01 Fireflies AI: Not just a conference recorder, but a comprehensive team assistant

As one of the best AI conference recording tools, Fireflies AI has solid "foundations": It supports 69 languages, and the real - time transcription accuracy is up to 95%. Within a few minutes, it can complete the transcription, summarization, and extraction of action points of a conference.

But this is just its "starting point".

What really differentiates Fireflies from its competitors is its strong ecosystem connection ability and AI collaboration engine. It not only takes notes but revises the entire information and implementation flow of a team.

Fireflies can be seamlessly integrated with over 40 office tools such as Zoom, Google Meet, Teams, Salesforce, Slack, Notion, etc. It automatically extracts important action points, customer requirements, and technical decisions from conferences and synchronizes them with different project management systems.

In other words: It bridges the gap between "understanding" and "doing", so that conferences not only serve for recording but also directly drive implementation.

For example:

The customer requirements put forward in a conference are automatically created as tasks in Slack, and the responsible persons are notified.

The technical solutions discussed in the team are automatically written as reminders in Notion by Fireflies.

The details of sales support can be entered into the CRM system with a single click.

In addition, Fireflies has access to the search engine engine of Perplexity and can thus search the Internet in real - time. Users can ask during the conference: "Fireflies, what are the current market forecasts for AI conference assistants in the industry?" The system immediately generates an answer and truly brings Artificial Intelligence into the discussion.

Furthermore, Fireflies plans to launch a dialogue agent named "Talk to Fireflies" and develop an AI surrogate that can "attend conferences for you". This surrogate can even retrieve contextual information from your previous conferences and automatically supplement missing information.

Currently, Fireflies has launched over 200 "AI agents" for different functions, each suitable for different roles and use - cases such as sales, marketing, personnel recruitment, operations and management, customer support, etc.

In the recruitment process, after the interview, the HR team can use Fireflies' AI applications (e.g., "Culture - fit Extractor" and "Candidate Feedback Aggregator") to generate evaluation cards for candidates.

In the marketing field, the product launch planning app can develop a product launch strategy, and the event performance evaluation app can extract important insights from event discussions.

In the sales field, the sales app can extract detailed information such as the budget and schedule of the sales team.

In the operations field, Fireflies' AI application (e.g., Standup Notes) can capture action points and obstacles, and the goal - tracking tracker can monitor progress.

In the customer service field, the agent performance feedback generator can generate insights into the performance of customer service staff and give consulting recommendations.

The underlying logic is clear: Each department has its own language and way of working, and Fireflies understands these "departmental dialects" through customized agents and helps you automatically execute tasks after the conference.

While AI conference notes are increasingly involved in "internal tool competition", Fireflies has obviously exceeded the limits of simple recording and is striving to become a real intelligent collaboration center that "understands human language and drives teams forward".

02 From Tool to Center: The Way Out for AI Notes

The scribe is an old profession that aims to record things.

In ancient times, this profession also had the functions of a secretary and administrative staff. It recorded orders for royal houses and temples, organized daily affairs, and later also worked in fields such as commerce and law, where a lot of language communication was required.

Today, the need for recording can be found everywhere in work and life, for example, in the fields of medicine, law, and government. As shown in the figure below, in the US transcription market, there are about 100,000 scribes in the medical field alone, which is one - tenth of the total number of doctors. Studies show that the volume of the US transcription market was 21.6 billion US dollars in 2020 and rose to 2.38 billion US dollars in 2021.

With the rise of Artificial Intelligence, this old industry, which was traditionally driven by human labor, has been completely overturned. The author has noticed that in addition to Fireflies.ai, a series of AI conference notes start - ups have also emerged overseas. They have not only achieved remarkable growth but also raised a lot of capital.

How do these AI conference notes companies manage to bypass the applications of large enterprises and find their own differentiated path in the fierce competition?

Granola, the "second brain" with a valuation of 250 million US dollars

Granola is one of the most exciting newcomers in this field.

According to the figures, the number of its users grows by 10% weekly, and the monthly retention rate is 70%. On May 14th, Granola announced that it had completed a Series B financing of 43 million US dollars and brought its valuation to 250 million US dollars.

Granola has chosen the frequent conference scenario and made many innovations at the product and interaction levels.

Granola sees itself as the "second brain" of users.

In contrast to most AI note - taking products that directly generate corresponding conference records, Granola chooses to cooperate with users. That is, users can record notes or spontaneous ideas at any time, and Granola captures and transcribes the conversation content in real - time. After the conference, it will further improve and enrich the transcribed content based on your notes.

Essentially, Granola has made a good division of labor between users and AI. Humans note down what comes to mind, and the AI notes down what is said in the conference. In the end, the Large Language Model combines both.

This approach fits better with real - world conference scenarios and can resonate more with users. This design precisely meets the need for "mixed thinking" in real conferences and gradually transforms Granola from a note - taking tool into a lightweight collaboration platform.

Notion, the Lego - like AI note - taking platform with a valuation of over one billion US dollars

Compared to Granola, which starts with the conference scenario, Notion has chosen a more difficult path - it wants to become a "general work platform".

Notion wants to assemble the functions you need with "building blocks". Documents, tasks, databases, and even AI assistants are all modular blocks. Users can assemble their work processes like building with Lego.

Specifically, Notion is characterized by the following features:

① Free switching: The platform integrates a variety of AI functions, so users don't have to switch back and forth between the platform and AI tools.

② Free text movement and functional modularity: Notion allows copying content between different editor tools. It can convert functions such as documents, databases, and tables into block - like modules. Users can freely create what they want. The operation is flexible, and the expandability is strong.

This reflects Notion's Lego - like product philosophy. It builds its software system on a block - based rather than an application - based method.

But the other aspect of the platform approach is that there is no natural industry barrier. Therefore, it has to create space for itself through user acceptance, product freedom, and the depth of collaboration.

So far, at least, Notion has done well. Thanks to its strong AI functions and convenient flexibility, Notion has increased the number of its users from 1 million to 100 million in over three years and raised its valuation to over one billion US dollars. Its investors include top institutions such as First Round, Index, and Sequoia.

Abridge: Specialization in Medicine and as an AI Recording Assistant for Doctors

Abridge has chosen a more specialized but promising direction: the medical scenario.

In contrast to lightweight products for private doctors, Abridge directly cooperates with US healthcare systems and becomes part of the medical information system (EMR). It has access to over 100 well - known institutions such as the Mayo Clinic, Duke University, and Yale University and covers thousands of clinical doctors.

Its technological foundation is completely tailored to medicine - its own large speech recognition model, which has been trained and fine - tuned for dozens of languages and over a hundred professional fields to ensure that the notes contain correct medical terminology, have a clear structure, and meet the writing norms of doctors.

Supported by this specific model, Abridge supports accurate summarization and medical terminology in over 50 professional fields and can recognize, understand, and accurately record notes in over 14 languages, ensuring the completeness and high accuracy of structured clinical note drafts.

In February this year, the AI medical note - taking company Abridge completed a financing of 250 million US dollars, and its latest valuation is an impressive 2.75 billion US dollars. The leadership was taken over by the well - known Silicon Valley venture capitalist, Elad Gil. Given the annual recurring revenue (ARR) of only 50 million US dollars at the end of 2024, the EV/ARR ratio is an impressive 55, even higher than that of companies with Large Language Models.

These AI notes, which are deeply integrated into professional work processes, also include players such as Suki AI, Nabla, and Freed AI and revise the generation logic of traditional medical documents.

03 Summary

Granola, Notion, and Abridge represent three development paths for AI note - taking products:

Start with conferences and become the team's second brain: like Granola, which emphasizes the cooperation between humans and AI.

Build a platform and become a general work center: like Notion, which emphasizes the Lego - like combination and ecosystem integration.

Specialize in a vertical industry and revise the professional scenario: like Abridge, which emphasizes data structuring and system integration.

Their commonalities are:

1) They all move from "recording" to "understanding". They are not satisfied with simple language transcription or conference records but understand the users' language through AI, extract structured content, recognize the intentions behind actions, and transform from information - recording tools into intelligent knowledge assistants.

2) They all build "larger containers". They are not "isolated tools" but seamlessly integrate into the users' existing systems through plugins, APIs, or native building blocks and become part of the core processes of the organization.

3) They all base on the product philosophy of "role understanding". All three have recognized that different user roles have different languages, tasks, and ways of thinking. Therefore, they all emphasize "design for the role".

Put simply, the common direction of these AI notes is that Artificial Intelligence no longer stays at the recording level but penetrates deeply into the "nervous system" of the organization and becomes the infrastructure for decision - making and collaboration.

This article is from the WeChat account "Raven Intelligence Talk"