Baloise CH

If your team depends on you for every answer, you’re managing. If they grow beyond you, you’re leading. 🚀 Empower. Delegate. Develop. #LeadershipDevelopment #Empowerment #TeamBuilding #Mentorship #BusinessGrowth

SF gets a bad rap for early dinners. Last night, our table was still going past 10! Filet mignon, good wine, and an enticing conversation on how coding agents are changing the workplace. This was Runtime (YC P26) 's second dinner, after our first one in NYC a couple of weeks ago. Same format, different city. Product and Eng leaders from companies like Salesforce, Waymo, Walmart, and some fast-growing scaleups. Here are some topics that came up: - Coding is the new baseline for PMs. Product leaders are asking for it, which means interviews for PMs are changing too. - Engineering leaders are drowning in bad PRs. A hot take was that opening the PR and fixing it yourself is faster than explaining why the implementation is wrong. - The line between engineering, product, and design is blurry but history tells us specialization will be needed. Time will tell how we'll need to re-specialize. Two dinners in, and it is becoming clearer that human connection is more important than ever. Reminiscent of how artists sat in Paris cafes in the early 1900s to talk about Cubism and the evolution of culture :) We'll be making more of these in the next weeks in both SF and NYC. Text me if you want in on the next one. Thanks for the great evening: Fawn Qiu Jason Edward Durrett Jeff Ammons Joe Constantino Jeremy Glassenberg Paul Cothenet Carlos Volante

This is the mindset shift: how do I prep the infrastructure for what's coming? Not just giving people assistants. Not just making faster horses. But prepping the organizational data and infrastructure so that when AI needs to work, it has everything it needs. That's what the customer context model is about. That's top-down thinking. How do I prep the org for AI — not just for today, but for what's going to happen next? That's a fundamentally different approach than buying an assistant and hoping the metrics move. #mosaicai #b2bsupport

People will realize the problem fast when the bill comes in. As more people use Claude internally, they'll say — where are all these credits going? This is the problem we're solving with our Customer Context Model at Mosaic AI. It has to happen in an agentic way because you're dealing with massive amounts of unstructured data — tickets, calls, emails, CRM entries, knowledge articles. The AI needs to go over all of it, extract and build structured entities you can work with. But you need to optimize for cost. If you have 100,000 tickets and run the most expensive model, you're looking at thousands of dollars. So you build that ETL with efficient models in a cascade. Summarize first. Then compute something else. Then build on that. Check accuracy at each step. It's pretty much like an art — prepping your company data for AI. And people are underestimating it. They think: just index everything, the magic will happen. OpenAI is magical. Claude is magical. No — they'll just ruin your credits when your users start asking hard questions. And believe me, they will. #mosaicai #b2bsupport