An engineering model where teams route through shared, centrally-built infrastructure and standard patterns ("paved paths") instead of each building a bespoke local stack. Elizabeth Stone (Netflix CPTO) frames this as increasingly necessary once AI agents — not just humans — touch more of the system: without commonly built source-of-truth data access, guardrails, and testing, speed outruns accountability.
Stone frames paved paths in AI-era terms with two purposes: leverage and safety. On leverage, a good internal platform should get roughly 80% of teams to a working solution without each team re-deriving the same building blocks — the platform absorbs repeated work so a growing number of bets doesn't require rebuilding fundamentals each time. On safety, as more people (and AI agents) can now ship code, design, or analysis directly, Netflix encodes "what good looks like" — access/identity rules, security requirements, and how to handle source-of-truth data — into the shared platform itself rather than leaving each builder to work it out ad hoc. The paved path is what lets systems-thinkers-over-specialists scale speed without losing accountability.