Lusha's R&D leadership decided to run an AI-native engineering organisation, and picked the two numbers they would judge it by: velocity, how quickly a team gets work into production, and autonomy, whether a team can finish something end to end on its own.
Engineers were already using AI to write code. The larger opportunity sat in everything around the code - how a requirement becomes a specification, how design reaches the backlog, how a release goes out, how the bug list gets worked. Each of those is a modest gain by itself. Together they are where the ceiling moves.