A team can now produce a working service in an afternoon. That's a real improvement. Six months later, though, someone still has to answer the pager, update its dependencies and explain why it can access customer data.
Consider a small internal assistant that starts as a quick experiment. It acquires a database connection, an API token and a handful of enthusiastic users. Soon another team depends on it. The original build was cheap; the responsibility it created is now permanent until someone deliberately retires it.
The useful accounting question is simple. What does this service cost to live with? Include monitoring, security review, ownership, integration, incidents and the work required to change it safely. Generation speed tells you almost nothing about those costs.
Architecture can make that trade-off visible early. Give services clear owners. Make interfaces and permissions explicit. Automate checks for the boundaries that matter. A fast start is welcome, provided the organization can afford the long middle.