5 ideas
Before optimizing or automating any process, first question and delete every requirement, attributing each to a named person rather than a department. Only then simplify, accelerate, and finally automate — because automating a step that should have been deleted multiplies waste.
Building elaborate robotic systems before a process is proven creates brittle complexity that is harder to debug than human labor. The faster path to scale is often to insert humans temporarily, learn where the real bottlenecks are, and automate only what survives that scrutiny.
Every constraint or specification should be traceable to a specific individual, never to a vague organizational source like 'safety' or 'legal.' Anonymous requirements accumulate unchallenged because no one can be questioned about why they exist or whether they still apply.
Treat the rate of iteration — not the perfection of any single design — as the primary lever for improving output, since faster cycles compound learning. The cost of moving slowly to avoid mistakes usually exceeds the cost of the mistakes themselves.
In 2017–2018 Tesla's heavily automated Model 3 line at its Fremont, California factory stalled far below target, and Elon Musk publicly conceded that excessive automation had been a mistake and that humans were underrated. In June 2018, working from the factory floor, Tesla stripped out failing robotics and built a new general assembly line (GA4) under a tent in the parking lot in roughly three weeks. That push let Tesla reach about 5,000 Model 3s in a single week at the end of the quarter, the episode McNeil, then a Tesla president, uses to ground the five-step algorithm.