Yes. I do "agentic engineering," primarily using Cline as it allows me to gas-and-brake the AI and review what it's doing on a granular level. So, think pair programming but my #2 is an LLM. I routinely reject turns when a given model goes off into space. I also routinely make hot edits to its changes before advancing, several times per day.
You can use these tools wisely without letting it run unverified carelessly.