AI is rewriting the future of knowledge work, and the transformation is happening faster than most realize. At our recent AiSummit, we hosted a compelling fireside chat between Hao Sang, Startups Lead at OpenAI, and David Shim, CEO of Read AI, diving deep into how artificial intelligence is revolutionizing the way we work, collaborate, and scale innovation.
The conversation revealed bold insights about enterprise AI adoption, the evolution of intelligent tools, and what's next for startups building on cutting-edge technology. Here's what emerged from their discussion.
From Price Prediction to Meeting Intelligence: David Shim's AI Journey
David Shim's first encounter with AI wasn't in a Silicon Valley boardroom — it was through a company called Farecast in the early 2000s. The startup used machine learning to predict airline ticket prices by analyzing disparate data sources: weather patterns, pricing history, carrier information, and load factors.
"Oren Etzioni thought he could reverse engineer airline price tickets," Shim explained. "He built a model that got so good they started building an insurance market against it."
This early experience with data-driven prediction models laid the foundation for Shim's vision at Read AI, where the company has raised $83M through Series B funding and now serves over 3M active users.
The Evolution of Meeting Intelligence: Beyond Basic Transcription
Read AI started with a contrarian approach to meeting analysis. Instead of focusing solely on what was said, the company built foundational models around sentiment and engagement — analyzing how people reacted to the words being spoken.
Key insights from Read AI's approach:
- Reaction matters as much as content: When someone nods during a pricing discussion, that emphasis gets factored into the AI's analysis
- Visual cues inform meaning: Frowning during delivery timeline discussions signals important concerns that get highlighted in summaries
- Actionable feedback beats raw data: Users don't want to know they're "doing a horrible job" — they want solutions and recommendations
"People start to freak out when you tell them they're crashing," Shim noted. "You gotta help me fix this problem."
Sales AGI: Predicting Deal Progression with AI
Read AI's latest innovation, Sales AGI, represents the next evolution beyond meeting summarization. Launched just 30 days before the AiSummit, the tool has already driven over $10.6M in deal recommendations.
The system works by:
- Analyzing meeting transcripts, emails, and Slack conversations
- Predicting when opportunities will progress (25% to 50%, 50% to 75%)
- Identifying deals likely to close or stall based on meeting quality
- Providing recommendations 2-3 days ahead of human sales updates
This creates unprecedented transparency in sales pipelines, with 75% of Fortune 500 companies already using Read AI's platform.
The Multiplayer Search Revolution
Shim identified a critical shift happening in AI: the move from individual summarization to "multiplayer search." This evolution promises to unlock collective intelligence across teams.
The progression looks like this:
- Summarization era: AI summarizes individual emails, messages, and meetings
- Search phase: Users query AI for specific information and memo generation
- Multiplayer search: Collective data from multiple team members improves everyone's results
"If you've got six salespeople doing five calls every single day — that's 180 calls a week," Shim explained. "You get common questions and successful responses, but that information stays siloed to individuals."
Multiplayer search changes this by surfacing what worked for colleagues, automatically sharing successful strategies across teams.
The TikTok-ification of Enterprise Decision Making
Perhaps the most provocative insight from the conversation was Shim's prediction about the future of enterprise software: "It's going to be more TikTok meets Tinder when it comes to decision making."
Instead of complex searches and prompt engineering, enterprise users will get simple, swipe-based decisions:
- AI presents recommendations based on data analysis
- Users swipe right to approve, left to reject
- Complex background analysis gets distilled into simple choices
Shim pointed to a Mag Seven company with 1,000 Read AI licenses and 200 product managers conducting 1,000 customer interviews weekly. Their AI system can now identify customers who complained about specific features, determine if they actually canceled, and recommend product roadmap decisions — all presented as simple approve/reject choices.
Enterprise AI Adoption: Data Portability and Digital Twins
The conversation revealed emerging trends in how enterprises are adopting AI:
Preference Portability: Employees want to carry their AI preferences between companies, similar to how they might insist on using Slack even in a Microsoft Teams environment.
Digital Twin Development: The more data you feed into AI systems, the better they become at mimicking your communication style and decision-making patterns.
Training Investment Protection: Two years of connected emails, messages, and prompts create significant value that employees don't want to lose when changing jobs.
Global AI Adoption: Emerging Markets Lead the Way
One of the most surprising insights came from Read AI's global usage patterns. Emerging markets are adopting AI faster than developed economies:
- Brazil: Zero to 250,000 active users in four months
- Colombia: Schools with 10,000 students all using Read AI for class analysis
- South Africa: 1-2% of the population actively using the product
"They are more likely to adopt AI than the developed world because they're open to finding the fastest solution," Shim observed. "They don't have legacy processes holding them back."
The Orchestration Layer: Humans Still in Control
Looking ahead, both leaders see a future where AI handles more tasks but humans remain the orchestration layer. Shim drew parallels to Two Sigma, the hedge fund that used AI for trading decisions while keeping humans as the final approval authority.
"You are still responsible for the actual output," he emphasized. "If you push something to a website with wrong text, you're responsible for it."
This human-in-the-loop approach will become critical as AI agents take on more complex tasks across organizations.
Key Takeaways for Startup Founders
The conversation offered several actionable insights for startups building with AI:
Start consumer-first: Read AI's 80%+ retention rate after 30 days comes from immediate value delivery — no lengthy enterprise sales cycles or complex implementations.
Solve jobs nobody wants: Meeting note-taking was a perfect target because it's universally disliked but necessary work.
Leverage falling costs: OpenAI's 80% price reduction enables more experimentation and full-context data storage.
Think global from day one: Emerging markets offer faster adoption rates and less legacy resistance.
Focus on outcomes, not features: Users want results, not more complexity to manage.
As AI continues reshaping knowledge work, the companies that succeed will be those that make intelligence accessible, actionable, and ultimately human-centered. The future belongs to tools that amplify human capability rather than replace human judgment.
~~~~~
AI Disclaimer: AI helped us summarize the key points from these videos, and our editorial team reviewed everything to make sure it's clear and correct.
