Humans of AI: Presented by WRITER Systems, Not Models with WRITER’s Head of AI Dan Bikel
Nov 12, 2025
Dan Bikel, Writer's Head of AI and a veteran in the field with three decades at leading labs, discusses the shift from model obsession to building reliable AI systems. He argues for focusing on business needs before diving into research and outlines his strategy of 'Scale, Speed, Specificity.' Bikel highlights the dangers of model unpredictability for enterprises and promotes the idea of an 'agentic age' where AI systems can act autonomously. He also shares insights on hiring and investment for successful AI operations.
AI Snips
Chapters
Transcript
Episode notes
Joining A Startup To Reverse Priorities
- Dan left big labs to join Writer and reverse his workflow: start with business problems, then design research to solve them.
- He describes this as a ‘‘complete reversal’’ from model-first work to business-first engineering.
Systems Over Models
- Shift from asking "what can this model do?" to "what does this business need?" reframes AI as a business asset.
- Building systems, not just models, separates experiments from operationalized value.
Randomness Harms Enterprise Reliability
- Models are deterministic but sampling adds randomness that helps human-like text and hurts enterprise reliability.
- For businesses, unpredictability scales into expensive mistakes, so observability matters more than charm.