The Conversation Is Still the Product

If you remember one thing: AI does not coach your people. You do. AI is the scaffolding — the prep notes, the pattern-matcher, the second opinion at 9pm before a hard 1:1. The moment your reports feel like they're being processed by a model instead of heard by a human, you've lost them. Track that line carefully.

What AI is genuinely good at here: surfacing things you forgot, drafting questions you wouldn't have thought to ask, and spotting trends across weeks of messy notes. What it's bad at: reading the room, holding silence, and noticing that someone's voice cracked when they said "fine."

Your job is to do the bad-at part better because the good-at part is now cheap.

Prep: Personalized Questions, Not a Template

Generic 1:1 templates ("What's blocking you? What's going well?") are how you get generic 1:1s. After about week three, your report is just answering the questions, not thinking.

Use AI to break the loop. Before each 1:1, paste in:

• Last 2-3 weeks of your private 1:1 notes for this person
• Their last self-review or any goals they wrote down
• Any recent Slack threads, PR comments, or project updates that involve them (redact anything sensitive — see [[chapter-11]])

Then prompt:

You are helping me prep for a 30-minute 1:1 with [name], who is a
[role] on my team. Based on the notes below, draft 5 candidate
questions for this week's conversation. Rules:

- No generic check-in questions ("how's it going").
- Each question should reference something specific from the notes.
- Mix: 1 about growth, 1 about a recent struggle or stuck point,
  1 about an opinion they hold strongly, 1 about energy or motivation,
  1 wildcard.
- Flag any patterns you see across weeks that I might be missing.

Notes: [paste]

You will not use all five. You'll pick one or two that land, and the rest will jog your memory. The "patterns you might be missing" line is the real value — that's the thing your report has mentioned three times that you didn't connect.

Tracking Growth Themes Across Time

Most managers run 1:1s like goldfish. Each week is a blank slate, and you're relying on your memory to notice that someone has been quietly anxious about scope for a month, or that they mentioned wanting to learn data work in January and you never followed up.

Build a simple per-report doc. Date-stamped, plain text, one section per 1:1. Two columns mentally: what they said, what you committed to.

Once a month, feed the whole doc to AI:

Below are my 1:1 notes with [name] from the past 8 weeks. Identify:

1. Recurring themes (what comes up more than once)
2. Goals or interests they've mentioned that I haven't followed up on
3. Shifts in tone or energy across the weeks
4. Commitments I made that don't appear to be closed

Be specific. Quote the notes.

Notes: [paste]

This is the move. The patterns are in your notes already — you just can't see them through the week-by-week fog. AI sees them in 30 seconds and tells you the thing you should have noticed.

Two warnings. First, keep these notes private — they're your working memory, not HR documents, and your report should know they exist if asked. Second, don't pipe sensitive notes through random consumer tools; use whatever's approved for confidential data in your org.

Drafting Coaching Plans You'll Actually Use

When someone is stuck — underperforming, plateauing, or just bored — the temptation is to either over-plan (a 12-week development framework nobody opens twice) or under-plan ("let's just see how next month goes").

AI is useful for the middle path. Give it the context and ask for a draft, then ruthlessly cut.

[Name] is a [role] who has been [specific situation: e.g., "shipping
on time but disengaged in meetings for 6 weeks"]. Their stated goal
is [X]. My read is [your hypothesis].

Draft a 6-week coaching plan with:
- 2-3 concrete behaviors I want to see change
- One thing I will do differently as their manager
- A check-in cadence
- What "this isn't working" looks like at the 6-week mark

Keep it to one page. No frameworks. No jargon.

The "what isn't working looks like" line forces honesty. Most coaching plans fail because nobody defined the failure state. If you can't name it, you'll let things drift for another quarter.

Then — and this is the part people skip — share the plan with your report. Not as a verdict. As a draft. "Here's what I'm thinking. What did I get wrong?" That conversation is the coaching.

What to Never Outsource

A short list, because it matters:

• Giving the feedback itself. AI can rehearse, you deliver. If you read off a screen, you've failed.
• Praise. Generic AI-flavored praise is worse than no praise. People can smell it. If you want to recognize someone, do it in your own words, even if they're awkward.
• Hard news. Performance issues, role changes, layoffs. AI helps you think through the words and anticipate reactions ([[chapter-6]] goes deep on this). It does not write the script you read.
• Listening. Put the laptop down. Take notes after, not during, when the conversation matters. Your report needs to feel like the most important thing in the room, because for 30 minutes they are.

If you want to go deeper on coaching mechanics generally, the AI for Managers Playbook course walks through frameworks you can adapt. For sharpening how you write feedback before it leaves your head, AI Writing & Content Creation is a faster read than you'd think.

The Weekly Habit

Fifteen minutes, once a week, before your 1:1 block:

1. Open each report's notes doc.
2. Paste into your AI tool with the prep prompt above.
3. Pick one question per person that you actually care about hearing the answer to.
4. Close the laptop.

That's the whole system. The model does the pattern-matching. You do the part that builds the human.