He saved OpenAI, invented the “Like” button, and built Google Maps: Bret Taylor (Sierra)
Sierra charges a pre-negotiated rate per resolved call, saving customers about $15 on a typical $10 to $20 call. Customers see 50 to 90% of service interactions fully automated, and one engineer root-causes every bad Cursor output by fixing its missing context.
Tips
- Price AI agents on outcomes, not tokens or usage▶ 1:07:42Sierra charges a pre-negotiated rate per resolution, saving the customer the cost of a call that typically runs $10 to $20, and some sales agents are paid a commission.
- Stack an AI reviewer on top of an AI generator▶ 1:11:02Two 90% accurate layers combine to roughly 99% accuracy.
- Root cause every bad AI-generated line by fixing its context▶ 1:11:40Sierra has an engineer who works exclusively on the model context protocol server serving its Cursor instance.
- Skip building frontier models; build applied agents instead▶ 55:00Taylor expects orchestrating agents to be easy in three or four years, so the product and workflow become the differentiator.
- Sell measurable autonomous jobs instead of vague productivity gains▶ 1:03:03








