Startup Stories – Mixergy
Episode #2280: Read.ai is adding 50k users per day
Host: Andrew Warner
Guest: David Shim, Founder of Read.ai
Date: September 15, 2025
Overview
In this episode, Andrew Warner interviews David Shim, founder of Read.ai—a rapidly growing AI meeting note-taker. They explore how Read.ai stands out in a crowded market, the nuances of AI-powered engagement and sentiment analysis, the evolution of meeting productivity tools, and the surprising directions in which users are taking the product. David shares entrepreneurial lessons from building multiple startups, why specialization in AI tools may not be sustainable, and actionable insights for founders building in AI today.
Main Discussion Points & Insights
1. The Explosive Growth of Read.ai
- User Base & Growth:
- Read.ai boasts millions of users, adding about 50,000 new accounts daily.
"That's a run rate of a million plus on a monthly basis, 12 million annually." (01:09)
- Growth has been driven by viral adoption: people encounter Read in meetings, receive its notes, and sign up.
- Read.ai boasts millions of users, adding about 50,000 new accounts daily.
2. What Differentiates Read.ai in an AI-Crowded Marketplace
- Table Stakes vs. Differentiation:
- While transcription and summarization are baseline features, Read.ai focuses on real-time sentiment and engagement analysis.
"It's very easy to build a basic model... For us, what's really resonated is our ability to actually measure sentiment and engagement in real time and then apply it to the text." (01:45)
- While transcription and summarization are baseline features, Read.ai focuses on real-time sentiment and engagement analysis.
- Narration Layer:
- Not just who said what, but how people react—boredom, engagement, sentiment shifts—becomes part of the meeting summary.
"That narration layer is that additional context... that goes in and says, this is compelling versus this isn't interesting." (02:16)
- Not just who said what, but how people react—boredom, engagement, sentiment shifts—becomes part of the meeting summary.
- Multi-modal Input:
- Read tracks head orientation and subtle cues—not facial recognition—to infer participant engagement.
"If you're looking at the camera straight ahead... if you look over this way and go to a fixed position... We have a model that says that's a second screen." (03:34)
- Read tracks head orientation and subtle cues—not facial recognition—to infer participant engagement.
3. Retention, Effectiveness, and User Needs
- Retention Metrics:
- Read.ai reaches 80% retention after 30 days, indicating strong product-market fit.
"If you use our product in a meeting and you get the reports, we see that the retention is 80% after 30 days." (06:12)
- Read.ai reaches 80% retention after 30 days, indicating strong product-market fit.
- Why Users Choose Read:
- It's not about visible analytics, but about smarter, more actionable summaries.
"They actually don't want to see that on the back end. They want the notes to actually take that into account." (05:21)
- It's not about visible analytics, but about smarter, more actionable summaries.
4. Generalist vs. Specialist AI Tools
- No Siloed SaaS:
- Unlike traditional SaaS, AI tools should adjust dynamically by context and role rather than being niche products.
"In this AI world, you don't need [separate tools]. I think that one AI solution needs to go in and say how do I tackle this for an entire organization." (09:07)
- Centralized intelligence enables company-wide value—sales, product, support can all extract insights.
"So you want one solution across an organization that could be that storage of intelligence... that everyone can tap into." (09:32)
- Unlike traditional SaaS, AI tools should adjust dynamically by context and role rather than being niche products.
- AI Personalization:
- The AI recognizes context (e.g., sales call vs. internal meeting) and tailors summaries and recommendations accordingly.
"If Andrew's doing a meeting and he's doing an interview, the AI will know that it's an interview and adjust to Andrew's style." (11:47)
- The AI recognizes context (e.g., sales call vs. internal meeting) and tailors summaries and recommendations accordingly.
5. Origin Story and Unexpected Market Applications
- Founding Inspiration:
- The idea struck David during COVID, while working remotely from Cabo, seeing wasted time and disengagement in video calls.
"I could realize, should I even be on this call or not?... I started to do some math. I was like, this is really expensive for how many people aren't paying attention." (12:49)
- The idea struck David during COVID, while working remotely from Cabo, seeing wasted time and disengagement in video calls.
- Beyond Sales:
- Early use was expected in sales, but surprising real-world stories emerged—such as helping users with early onset dementia remember family conversations, or healthcare compliance and field notes.
"I have early onset dementia... I want to remember the things we talked about so they don't feel bad." (15:08) "Healthcare workers... are recording the conversation with their consent... and now they can go back [to analyze sentiment and engagement]." (16:38)
- Early use was expected in sales, but surprising real-world stories emerged—such as helping users with early onset dementia remember family conversations, or healthcare compliance and field notes.
6. AI as Real-Time Coach & Post-Meeting Analysis
- Real-Time Feedback:
- Read.ai initially focused on real-time coaching for sellers but found post-game, after-call coaching more effective due to cognitive overload.
"What we found was it is valuable for that niche use case. But a lot of times it becomes cognitive overload..." (18:22)
- Read.ai initially focused on real-time coaching for sellers but found post-game, after-call coaching more effective due to cognitive overload.
- Tangible Improvement:
- Coaching recommendations (e.g., pace, interruptions, language) are sent post-call, allowing users to self-correct without manager pressure.
"They're not defensive because it's the AI telling you, it's not your manager telling you..." (19:25)
- Coaching recommendations (e.g., pace, interruptions, language) are sent post-call, allowing users to self-correct without manager pressure.
7. Building and Funding Read.ai
- Product Development:
- Early models were trained in-house due to lack of public data; actors were recruited, but subtle, realistic cues proved most valuable.
- Mechanical Turk and manual labeling helped establish ground truth.
"We built them in house for us... really expensive. When we started, we were lucky enough to raise a large round." (26:14)
- AI/ML Background:
- David leveraged prior experience in machine learning (Faircast airfare prediction) and training models using weak and strong signals (location intelligence at Placed).
"So it really is this concept of weak and strong signals to recommend or to predict where you are." (28:24)
- David leveraged prior experience in machine learning (Faircast airfare prediction) and training models using weak and strong signals (location intelligence at Placed).
8. Product Ecosystem and Platform Risk
- Integrations and Contextual Intelligence:
- Read.ai connects with Gmail, Outlook, Google Drive, Notion, Jira, and more, unifying action items across meetings, emails, and documentation.
"Those action items in a silo of a meeting... But in an email if you followed up... now that action item is complete..." (30:54)
- Read.ai connects with Gmail, Outlook, Google Drive, Notion, Jira, and more, unifying action items across meetings, emails, and documentation.
- Platform Risk Mitigation:
- Looming competition from Zoom, Google, Microsoft Copilot, etc., isn't a major threat because cross-platform users don't want fragmented data and find native solutions too basic.
"There's not much loyalty on platforms. For our users, on a weekly basis, we see that they use more than one platform." (36:06) "We've seen more than a 15x increase in Microsoft Team meetings that we've measured since Microsoft Copilot launched..." (36:06)
- Looming competition from Zoom, Google, Microsoft Copilot, etc., isn't a major threat because cross-platform users don't want fragmented data and find native solutions too basic.
9. User Feedback, Metrics, and Unexpected Growth
- Global Adoption:
- Organic, viral growth in unexpected markets like Brazil's Portuguese dialect and Colombian universities (1–2% of national student population).
"Every single class now in Colombian universities are recorded with Read... and they built this storage of intelligence for the university..." (39:29)
- Organic, viral growth in unexpected markets like Brazil's Portuguese dialect and Colombian universities (1–2% of national student population).
- Rapid Feature Launching:
- Continuous iteration based on in-app metrics, A/B tests, and user feedback, empowered by tools like Statsig.
"We use a company that recently was acquired, Statsig from OpenAI... they let us test different things out." (39:29)
- Continuous iteration based on in-app metrics, A/B tests, and user feedback, empowered by tools like Statsig.
10. Advice for AI Startup Founders
- Launch Fast, Iterate Often:
- Tolerance for imperfection is higher than ever; ship quickly, fix, and improve.
"If you launch a product and you're not embarrassed by it, you're not launching it in a timely manner... but it has never been more true when it comes to AI." (42:29)
- Tolerance for imperfection is higher than ever; ship quickly, fix, and improve.
- Distribution Is Key:
- Many great ideas fail due to lack of adoption and awareness, not technological shortcomings.
"Half the companies [I've invested in] have been wound down... it's not that the technology wasn't great. It's that they didn't have distribution." (47:37)
- Many great ideas fail due to lack of adoption and awareness, not technological shortcomings.
- Look for “Table Stakes” Opportunities:
- Areas such as video authenticity (AI-generated vs. real) are untapped, high-potential markets.
"I do believe in the amount of the next year or two, it is going to be almost impossible to figure out is this AI generated or not..." (46:05)
- Areas such as video authenticity (AI-generated vs. real) are untapped, high-potential markets.
Notable Quotes & Memorable Moments
-
On Standing Out in AI:
"There's just a lot of noise in the market right now." (00:34)
"For us, what's really resonated is our ability to actually measure sentiment and engagement in real time." (01:45) -
On Product Design:
"What we found was people were incredibly surprised to say, Oh, wow, I was actually talking really fast. I can barely understand myself." (19:27)
-
On Organizational Productivity:
"We are the system of record for meetings. Where we become is the system of record for productivity." (09:32)
-
On AI’s Evolution in Meetings:
"We classify meeting types when that data comes in... all of that context... can be applied into the meeting summary." (11:47)
-
On Real-World User Impact:
"I have early onset dementia... I want to remember the things we talked about so they don't feel bad." (15:08)
-
On Competing with Platforms:
"For our users, on a weekly basis, we see that they use more than one platform... They don't talk with each other... So I'm never going to look at this and it's a waste of time." (36:06)
-
On Launching Fast in AI:
"If you launch a product and you're not embarrassed by it, you're not launching it in a timely manner, those things are all true, but it has never been more true when it comes to AI." (42:29)
Key Timestamps
- [01:09] – Read.ai’s current user growth and scale
- [01:45] – Differentiation: Sentiment/engagement analysis vs. basic transcription
- [03:34] – How engagement is measured with head orientation and movement
- [06:12] – 80% retention after 30 days; validating product-market fit
- [09:07] – Why meeting tools won’t silo by role; value of unified intelligence
- [12:49] – David’s founding story: inspiration in Cabo, seeing disengagement in meetings
- [15:08] – Case study: unexpected impact in healthcare, memory impairment
- [18:22] – Real-time coaching vs. post-meeting recommendations
- [26:14] – Building foundational AI models; funding and data creation
- [30:54] – Integrating email, docs, CRM to close the action item loop
- [36:06] – Platform risk analysis and cross-platform usage patterns
- [39:29] – Feature iteration, A/B testing, international viral adoption
- [42:29] – Actionable tips for founders: launch fast, iterate, focus on distribution
- [46:05] – The coming wave: AI-generated media detection and authenticity seals
Final Thoughts
This episode shines a light on how modern AI startups can differentiate and scale in competitive markets. David Shim’s journey demonstrates the power of listening to actual user behaviors, the value of core technological differentiation, and the necessity of rapid iteration and distribution. For founders, Read.ai’s playbook and real-world lessons offer a blueprint for building lasting, impactful products in AI’s fast-evolving landscape.