Audit your own work for AI.

1. List the tasks you repeat.For one working week, note anything you do more than twice. Not a wish list of things you should automate. An honest log of what you actually did.
2. Time them honestly.Include the thinking and the checking, not just the typing. A task that "takes ten minutes" often takes ten minutes plus the fifteen you spend fixing it afterward.
3. Score each one on three axes.How often it happens, how much it drains you, and how easy it is to check whether the output is right. That third axis is the one people skip, and it decides whether AI can help at all: if you cannot tell good output from bad quickly, AI will not save you time, it will just move the work to checking.
4. Pick one.Not five. One task, tried properly, teaches you more than five tasks tried badly.
5. Run it through the calculator, then test it for real.Enter your honest numbers and see whether they support trying it. Then time one real attempt with AI and compare it to what you estimated.

Step 3, worked out. Score each candidate task 1 to 5 on each axis. Add them up. The highest total is where to start.

Task | How often (1-5) | How much it drains you (1-5) | How easy to check (1-5) | Total
     |                 |                              |                         |
     |                 |                              |                         |
     |                 |                              |                         |

If every task scores low, or the calculator says a task would cost more time than it saves, that is useful information, not a failed audit. It usually means the checking step is the hard part, and no tool changes that until the task itself is simplified.

A negative number from the calculator is not a reason to give up on AI. It is a reason to pick a different task, or a smaller piece of this one, before spending more time on it.