Why your AI output feels inconsistent -

The assistant is not always the issue. Sometimes the role is undefined.

If an employee had no job description, no quality standard, no audience, and no finish line, you would expect inconsistent work.

That is how many people use AI.

They open a chat and ask for whatever is on their mind. Sometimes writing. Sometimes research. Sometimes strategy. Sometimes editing. The assistant can technically do all of it, but that does not mean it knows what good means in the current context.

The fix is simple: give the assistant a job description before the task.

Use six lines:

1.Role: what it is responsible for.
2. Audience: who the work is for.
3. Outcome: what should change.
4. Standard: what good looks like.
5. Boundary: what it should challenge.
6. Finish line: when the task is done.
This turns a vague request into an operating instruction.

Instead of asking for content, ask for a specific kind of content for a specific reader with a specific standard.

Instead of asking for ideas, ask for ideas that can be tested manually, tied to a buyer, and killed if there is no response.

AI is more useful when it knows the job.

The job description does not need to be long. It just needs to be clear.

More Than Lucky — Modern AI and systems for work, business, and life.