Building Generative AI features
summary
Startups describe concrete wins: Vital uses Amazon Nova Micro and Anthropic Claude Sonnet 3.5 to categorize ER notes with 95% accuracy, finding hundreds of incidental issues at one Arizona site and generating $750,000 in radiology follow‑ups; Tappy cuts invoice processing cost 10x with multimodal image analysis; Sapia AI scales interviewing into 50+ languages and saves recruiters ~20 hours/week; Ask Nicely ships dynamic surveys with 46% adoption and 3× response rates. Practical design tactics: task focus, narrow scope, asynchronous work, and augment not automate.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 1:06Why generative AI is hard
- 3:35Key challenges with LLMs
- 5:42When to use generative AI
- 7:07Generative AI superpowers
- 16:45AI product design strategies
- 25:17Closing and key takeaways
related talks
27mBuilding AI applications: Successes and Lessons
17mGenerative AI: Keeping it real
45mThree Practical Ways to Use AI in Your Company
56mEnabling teams with AI
42mUX+AI: Mastering the UX of AI-Powered Products
49mHow to be an AI First company
23mGenerative AI use cases
01h 06mAI in the Trenches: Engineering Real-World Systems