Agentic AI insights from Silicon Valley: As Taiwan becomes the backbone of the AI revolution, Mosaic Venture Lab invests in startups that can harness Taiwan's unique advantages in this transformation. Here are the raw insights from Des Traynor (Intercom co-founder) and Ethan Kurzweil (Chemistry VC) on what it actually takes to survive and thrive in the AI era. Johnny Yu Ken Huang
Cross-Border Tech Innovation | Startup Investor via Taiwan CVC Fund | Automotive, Biotech & AI | US-Taiwan-Germany | Ex-Audi Innovation
𝐀𝐈 𝐀𝐠𝐞𝐧𝐭𝐬: Just got back from an fireside chat with Des Traynor (Intercom co-founder) and Ethan Kurzweil (Chemistry VC). The backstory here is wild. By 2022, Intercom was struggling hard. Five consecutive quarters of declining new revenue growth. Owen (the co-founder) had to return as CEO. Their plan? Compete directly with Zendesk because "our product is just basically better." That plan lasted 7 weeks. November 2022: Fergal (their AI director at the time, now Chief AI Officer) sends Des a short Slack message: "chat.openai.com - you should look at this." Des tries it. First question: "How do you install Intercom on iPhone?" ChatGPT nails it perfectly in 7 seconds. Better than any human support rep could do that fast. That weekend changed everything. By Sunday, Owen decided they were going all-in on AI instead of just having "nicer UI than Zendesk." 𝑇𝘩𝑒 𝑝𝑎𝑟𝑡 𝑡𝘩𝑎𝑡 𝘩𝑖𝑡 𝑚𝑒: Des said companies with healthy businesses struggle most to adapt to AI because "people change through inspiration or desperation, and we definitely had a healthy cocktail of both." What they actually built is insane: Everyone thinks Fin.AI is just another chatbot wrapper. It's actually 27 different subsystems - fatigue engines, disambiguators, retrieval systems. Des: "We've had to go really deep to nail this job." They completely flipped their product development process. It used to start with product managers making wireframes, beautiful designs. Now: "Can we reliably do this with AI 99 times out of 100?" Only after that passes do they think about UI. Des admitted: "People felt like we're losing our soul. We had to rip it all up and throw it away." 𝐐&𝐀 𝐡𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬: 𝐎𝐧 𝐨𝐮𝐭𝐜𝐨𝐦𝐞-𝐛𝐚𝐬𝐞𝐝 𝐩𝐫𝐢𝐜𝐢𝐧𝐠 (99¢ per resolved ticket): Des explained they were "displacing labor" so they should "price like human labor replacement." The math worked because average CS reps answer ~10 questions/day, so 25% automation at 99¢ was competitive. 𝐎𝐧 𝐂𝐑𝐌 𝐫𝐞𝐢𝐧𝐯𝐞𝐧𝐭𝐢𝐨𝐧: Ethan called current CRMs trash because you can't query them naturally. Want to ask "show me all customers that look like X"? Impossible. He sees this as a massive opportunity. 𝐎𝐧 𝐒𝐥𝐚𝐜𝐤: Des went hard: "Never has a product needed to be rebuilt and reconsidered in a post-AI world more than Slack." He thinks it's a $30B opportunity if someone attacks workplace communication with native AI. 𝐓𝐡𝐞𝐢𝐫 𝐟𝐚𝐯𝐨𝐫𝐢𝐭𝐞 𝐀𝐈 𝐭𝐨𝐨𝐥? Granola for meeting notes. When Intercom's security team banned it, employees revolted. Des got 150+ Slack messages demanding he "sort this xxx out." The uncomfortable insight: Des used this phrase that stuck with me: "If you're going to eat ***, you don't nibble." Sometimes you have to completely rebuild everything at once instead of gradual changes. For most companies, the AI reckoning is coming. Some will recognize it. Others won't even see what hit them. What's your industry's customer support moment going to be?