Name the skill, then the drill.

Can you name the skill each person on your team is building? “Be more senior” is not something anyone can practice on Thursday.

AI prepares the edges. You own the middle. The edge is a list of drills for one named skill. The middle is the skill you picked, and the standard you will actually hold.

Why development turns into tasks

A full week has no empty slot labeled growth. So the person gets more of the work they already know how to do. It looks like trust. It does not teach anything.

The miss happens at the start. If you cannot name the skill, you cannot tell a task from a rep. You also cannot tell the model what to design. It will hand you a course outline. You needed four weeks of practice on real work.

Name it so it can be practiced

A skill is a thing a person can do in front of you, badly at first.

• Too wide: communication, ownership, leadership.
• Narrow enough: write a design note a reviewer can challenge. Run a rollback without you in the room. Give a code review that says what to change and why.

One skill. Not a development plan with six themes. One, for the next four weeks.

You set the bar

Before anyone asks for drills, write the standard in a short paragraph you would actually say.

What good looks like. What is not good enough. One example of the difference, described, not performed as a story about a named person.

The model does not set this. If you skip it, every rep it invents will aim at a bar you have not agreed.

AI writes the reps

Paste the skill and the standard. Ask for four weeks. Each week is one rep, on work the team already has, small enough to finish, reviewed by you against the standard.

A useful drill sheet looks like this.

• Week 1. They do a small piece beside you. You mark it against the standard, in writing.
• Week 2. They do the next piece with less of you in it. Same standard.
• Week 3. They do it, and they write what they would change before you comment.
• Week 4. They do it on a real piece of work. You say whether it met the bar. Yes or no, and why.

You assign the rep. You review the work. The model does not score the person.

Weekly cadence

1. This week: name one skill for November. Write the standard in one paragraph.
2. Ask for four reps. Cut any rep that is a course, a blog post, or a task with no review.
3. Each week: one rep, on real work. You review it against the paragraph you wrote.
4. At the end of week 4: the bar was met, or it was not. Then you decide the next skill. Not the model.

Mistakes to avoid

• Naming a trait instead of a skill.
• Letting the model pick the skill from a job description.
• Accepting a four-week reading list as practice.
• Scoring the person in the tool. Score the work, in your words, against the standard.
• Pasting private performance notes in to “personalize” the drills.

Before the review

The four-week drill gives you dated evidence against a written standard. That is what a year-end review needs. Without a named skill and a standard, the review becomes adjectives. AI can sort that evidence. You decide the rating. The full process is in the guide: Should managers use AI to write performance reviews?

Tools

Whichever model the company already trusts, used to draft the reps. The sheet lives with you. Do not put the review of the person into a tool the company has not approved.

The four-week drill sheet, with the drill prompt on it, is a free one-page PDF.

New here? Read who writes these notes, browse all weekly notes, or get the free kit below.