Software engineering teams evaluate six primary constraints when adopting AI workflows: token budget, organizational readiness, goal definition, domain knowledge, verifiability, and code oversight