Prep Every 1:1 from Your Own Notes — Behind a Hard Privacy Boundary

Workflow area: Leadership Ops · Tier: Core · Hands-on time: ~30 minutes

The difference between a good 1:1 and a wasted one is usually memory: what you discussed last time, what you committed to, what pattern is forming across the last two months. AI is excellent at holding that memory — and dangerous at it, because your 1:1 notes are the most sensitive text you produce. One pasted paragraph of peer feedback or a health disclosure into the wrong tool is a trust breach you cannot walk back. The boundary list comes first; the prompts come second.

When to use — and when not to

• Use it for every recurring 1:1 with a direct reportthe compounding value is in continuity, and continuity is exactly what a busy leader's memory drops.
• Use it when you inherit a teamsynthesizing your first six weeks of notes per person surfaces the pattern you are too new to see.
• Use it before growth conversationsa brief built from their stated goals plus your observations produces better development questions than either alone.
• Skip it entirely on free or consumer AI tiers1:1 notes go only into an enterprise tenant with no-training guarantees; if you don't have one, this playbook waits until you do.
• Skip it for active HR situationsperformance management, complaints, accommodations, or anything a lawyer might one day read follows pb-22's rules, not this playbook.
• Skip the AI when the 1:1 needs to be human-shapeda report in distress needs your attention, not your agenda; close the brief and listen.

The recipe

1. 1.

   Write the never-enters-the-prompt list before your first brief. This is a real document, not a mental note. Minimum contents: peer feedback verbatim or attributable (paraphrase to unattributable themes only, and only with care), health and family disclosures, anything told to you in confidence, compensation details, unannounced org or role changes, immigration status, and anything about a third person that they would not want forwarded. Keep the list next to your notes template where you will see it every time you paste.

2. 2.

   Keep structured 1:1 notes worth synthesizing. After each 1:1, three lines in your private notes: topics and decisions, commitments (yours and theirs), observations. Sanitize as you write — "discussed a personal matter, following up in two weeks" instead of the matter itself. Notes written clean are notes you can safely process later.

3. 3.

   Before each 1:1, generate the brief. In your enterprise-tier Claude or ChatGPT:

   Copy

   You are helping me prepare a 1:1 with a direct report. Context:
   [ROLE, TENURE, CURRENT PROJECT — no name needed, use "R"].
   Their stated goals this half: [PASTE THEIR OWN WRITTEN GOALS].
   
   Here are my sanitized notes from our last [N] 1:1s:
   [PASTE NOTES]
   
   Produce:
   1. OPEN LOOPS — commitments either of us made that have no
      recorded closure, oldest first.
   2. PATTERNS — themes recurring across 3+ sessions, each with the
      note lines that evidence it. No armchair psychology: describe
      the pattern in the notes, not a theory about the person.
   3. GOAL CHECK — where recent topics connect to their stated
      goals, and any goal we have not touched in a month.
   4. THREE QUESTIONS — open questions I could ask, tied to the
      above. Questions, not scripts.
   
   Rules: use only what the notes support; write "nothing found"
   over inventing; do not rate, rank, or diagnose the person.
   

4. 4.

   Read the brief as input, not agenda. Good output looks like:

   Copy

   OPEN LOOPS
   - You committed to intro R to the platform lead (5 wks ago) —
     no closure recorded.
   PATTERNS
   - R has raised on-call load in 4 of the last 6 sessions, each
     time framed as "fine but" — see notes 6/12, 6/26, 7/10, 7/24.
   THREE QUESTIONS
   - "You've mentioned on-call four times — what would make it
     actually fine, not 'fine but'?"
   

   Pick the one or two items that matter; discard the rest. The report sets half the agenda — a fully pre-scripted 1:1 is your meeting, not theirs.

5. 5.

   Close the loop after the 1:1. Two minutes: log what happened against the open loops, add new commitments. This is what makes next month's brief accurate — the pipeline is only as good as the notes you feed it.

Make it yours

• Manager of 8+ reports: run the brief weekly as a batch, one chat per person (never one chat comparing people — cross-report synthesis is how briefs turn into rankings, and rankings built by AI are pb-21 territory with rules of their own).
• Skip-levels: same recipe, thinner notes, and one added rule — the brief must never contain anything their direct manager told you about them; skip-level trust dies the day it becomes a triangulation channel.
• Startup founder with no enterprise tier: keep the synthesis manual — reread your notes with the same four headings as a checklist. The structure is the value; the AI is the accelerant, and without a no-training tenant the accelerant is not worth the fuel.

Quality check

• The boundary list heldreread your last three pasted prompts; if anything on the never-enters list got through, tighten how you write notes, because the leak is upstream of the paste.
• Patterns cite note linesa "pattern" without evidence lines is the model narrating; delete it.
• No personality theories survivedthe brief describes what the notes say happened, never what kind of person R is; AI-generated character analysis of your reports is both wrong and corrosive.
• Open loops are actually openverify against your tracker before raising a "you never did X" that you did.
• The human conversation stayed primaryif you ever catch yourself trusting the brief over what the person says in the room, stop using briefs for a month.