The 1-on-1 starts in 12 minutes. You opened the doc once last week. The last entry says “career chat — TBD.” This tutorial replaces the “wing the 1-on-1 then forget the action item” cycle with a 10-minute prep workflow plus a same-day follow-up note: the AI builds the agenda from your last 2 weeks of context, drafts the topics worth raising, and writes the post-meeting note. It is written for both sides — managers running 4-8 reports, and reports who want their 1-on-1 to be more than a status check.
TL;DR
- A repeatable 5-prompt workflow: context pull, agenda draft, hidden topics, decision target, same-day follow-up. Steady-state prep time is about 10 minutes.
- Flip the no-training setting before you paste anything personal. As of June 2026, none of the consumer chatbots stop training on your conversations by default — you toggle it off, or use a temporary/incognito chat (details and table below).
- Tool matters far less than cadence. Gallup’s workplace research finds employees with regular manager meetings are roughly 3x as likely to be engaged; weekly 30-minute or biweekly 45-minute is the cadence most teams settle on.
- The AI handles summarizing and structuring. You keep the trust judgment, the personal questions, and the candor a model cannot fabricate.
Who this is for
Managers running 4-8 direct reports across functions, senior ICs who manage zero people but mentor 2-3 juniors, reports who want to extract more value from their 1-on-1s, and skip-level managers who do quarterly 1-on-1s with grandchildren on the org chart. It works best when both parties want the 1-on-1 to be more than a status check. It will not help when one side has already decided the 1-on-1 is overhead — the AI cannot fix a relationship problem.
When to reach for it
Weekly or biweekly recurring 1-on-1s, a quarterly career conversation, a skip-level you only do once a quarter, or after a hard week when you need to walk in clear about what to say. Also useful when you take over a new team and your first 1-on-1s with each report need to be more than introductions.
Flip the privacy setting first
1-on-1 prep contains the most sensitive text most people ever paste into a chatbot: names, performance issues, comp gripes, sometimes health. As of June 2026, the consumer tiers of ChatGPT, Claude, and Gemini all train on your chats by default — you have to opt out, or run the conversation in a mode that is exempt.
| Tool (as of June 2026) | One-time opt-out | Per-chat exempt mode | Default retention after opt-out |
|---|---|---|---|
| ChatGPT (Free/Go $8/Plus $20/Pro $100-200) | Settings → Data Controls → “Improve the model for everyone” → Off | Temporary Chat (not saved, not trained on) | ~30 days for safety review |
| Claude (Free/Pro $20/Max $100-200) | Settings → Privacy → “Help improve Claude” → Off | Incognito chat (ghost icon; never trained on, even if model improvement is on) | 30 days |
| Gemini (Free/AI Pro $19.99/AI Ultra $99.99) | Activity → “Keep Activity” → Off | Temporary chat | ~72 hours |
Two notes that catch people out:
- A paid plan is not a privacy plan. Claude Pro at $20 and ChatGPT Plus at $20 still train on your chats unless you toggle the setting. (Anthropic’s data-training page and OpenAI’s data controls both spell this out.)
- The only setups that bar training by contract rather than a toggle you might forget are the business tiers — ChatGPT Business/Team and Claude Team — which do not train on customer content. If your 1-on-1 prep touches comp or HR cases, use one of those or keep that detail offline.
Before you start
- Confirm the recurring slot. 30 minutes weekly or 45 minutes biweekly is the typical rhythm; less than 30 minutes is a status check, not a 1-on-1.
- Have a shared running doc — Notion, a Google Doc, or a Linear page — that both sides edit. Verbal-only 1-on-1s lose continuity by week 3.
- Know each report’s current top project and current top concern. If you cannot name them before the meeting, the meeting is going to be shallow.
- Decide which tool, and confirm its no-training setting per the table above.
- Block 10 minutes before the 1-on-1 starts — not the moment it starts.
Step by step
Save the five prompts below as reusable snippets. Per-person, they barely change. Replace [name] and the bracketed placeholders each time.
1. Pull context
Here is the last 2 weeks of context for my 1-on-1 with [name]:
[paste recent Slack threads, project status, last meeting note].
Summarize as: 3 things going well, 2 things stuck, 1 emerging risk.Read the summary. It usually surfaces something you forgot.
2. Draft the agenda
Draft a 30-minute 1-on-1 agenda for [name] given the context above.
Structure: 5 min check-in, 15 min topic-driven discussion,
5 min career and growth, 5 min open floor.
Suggest 2-3 specific topics for the topic block, ranked by urgency.Cut to your top 2 topics.
3. Surface the topics they won’t raise
What does this person probably want to raise but might not bring up
unprompted? Consider: project bandwidth, growth path, frustration with [X].
List 3, ranked by how likely silence on it leads to a bad outcome.Pick 1 to proactively invite into the room.
4. Name a decision to land
Write 1-2 specific decisions you want to leave the room with. “Decide whether to extend the sprint by a week” beats “discuss sprint pacing.” Without a decision target, 1-on-1s drift into venting. The AI can pressure-test it: Is this a real decision or a discussion topic in disguise?
5. Hold the meeting
Spend the first 5 minutes letting the other side set the agenda. The AI’s draft is a backup, not the script. The report (or the manager, from the report’s side) gets the first 5 minutes uninterrupted.
6. Same-day follow-up note
Here is what we covered: [paste 4 bullets].
Draft a 6-bullet follow-up: decisions made, action items with owners,
items deferred, next 1-on-1 focus. Keep it under 30 seconds to read.Send before end of day. If it reads like an AI summary, rewrite the action lines by hand so the other side recognizes the meeting.
7. Quarterly review
Read the last 12 follow-up notes together. The pattern to watch: the same topic raised 4 times unresolved is a structural problem, not a 1-on-1 problem.
Which model to use
Any current frontier chatbot handles this workflow; the differences that matter for 1-on-1 prep are context window (how much history you can paste) and how clean the no-training story is.
| Model (June 2026) | Plan | Why for 1-on-1 prep |
|---|---|---|
| Claude Sonnet 4.6 / Opus 4.7 | Free / Pro $20 / Max $100-200 | 1M-token context; incognito chat is exempt from training even with model improvement on — cleanest privacy default here |
| ChatGPT GPT-5.5 | Free / Go $8 / Plus $20 / Pro $100-200 | Temporary Chat exempts a session; Plus in-app context is ~320 pages, plenty for 2 weeks of threads |
| Gemini 3.1 Pro | Free / AI Pro $19.99 | 1M context and native Google Docs/Workspace pull if your running doc lives there |
For most people, the chatbot already paying for is fine. Paste 2 weeks of context, not 2 months — more history makes the summary vaguer, not sharper.
First-run exercise
- Pick your next real 1-on-1 — not a hypothetical one. Run the prep workflow 24 hours in advance the first time, even though the steady-state goal is 10 minutes prior.
- After the meeting, write the follow-up note within 2 hours. Same-day discipline matters more than format quality.
- For the second 1-on-1, change only one variable: a tighter context pull, a sharper decision target, or a different model.
- After 4 weeks of the routine, audit: are 1-on-1s producing decisions you can name? If not, the agenda is still a status check in disguise.
Quality check
- The agenda includes at least one topic neither side raised last week. New ground beats rehashing.
- One specific decision is named before the meeting starts. Decision-less 1-on-1s become venting sessions.
- The follow-up note has explicit owners on every action item. “We will look into X” is not an action item.
- Career and growth gets 5 minutes every meeting. Skipping it for 6 straight weeks tells the report their career does not matter to you.
- Privacy: no client name, comp detail, or personal health info gets pasted into a consumer chat without the no-training setting on or a temporary/incognito session.
- The other side could read the follow-up note and recognize the meeting.
How to reuse this workflow
- Keep the context-pull, agenda, hidden-topics, and follow-up prompts as 4 saved snippets. Per person, they barely change.
- Maintain a per-report folder with the running doc, weekly follow-ups, and quarterly themes. After 2 quarters, patterns are unmissable.
- Quarterly, run a meta-prompt:
Read 12 weeks of 1-on-1 notes with [name]. What 3 themes recur? Which were resolved? Which are still open?Use the output to seed the quarterly career conversation.
Recommended workflow
Context pull → agenda draft → hidden topics → decision to land → meeting (other side first) → same-day follow-up → quarterly retro. For the meeting-day half — recording, transcribing, extracting action items — the AI meeting summary tutorial covers it.
Common mistakes
- Skipping the prep because “we have a running doc.” A running doc is necessary but not sufficient; the prep surfaces what is missing from the doc.
- Letting the AI write the agenda verbatim. AI agendas are smooth and generic; real 1-on-1s need named tension.
- No follow-up note. The conversation evaporates by Tuesday and the next 1-on-1 starts from zero.
- Treating every 1-on-1 the same as the last. The same agenda 6 weeks in a row signals that no one is paying attention.
- Hiding the hard topic to keep the conversation pleasant. The unraised topic becomes a resignation in 90 days.
- Pasting sensitive HR or comp info into a default consumer chat. Toggle no-training on, use an incognito/temporary chat, or keep that info offline.
FAQ
- Manager or report — who should prep? Both. Prep is asymmetric: the report drives content, the manager drives structure. The AI helps both sides.
- How often should 1-on-1s be? Weekly 30 minutes for direct reports; biweekly 45 minutes for mature reports who run themselves; quarterly for skip-levels. Gallup’s research ties regular manager meetings to roughly 3x higher engagement, so the cadence matters more than the polish.
- Is it safe to paste 1-on-1 notes into ChatGPT or Claude? Only after you flip the no-training setting or use a temporary/incognito chat — and never with comp, HR-case, or health detail in a default consumer chat. See the privacy table above.
- Do I need a 1-on-1-specific tool like Lattice or 15Five? No. Lattice’s core modules start around $11/seat/month (with a ~$4,000 annual minimum) and 15Five starts around $4/seat/month, both as of June 2026 — useful at scale, but a shared Notion or Google Doc plus this workflow covers most teams. Discipline beats tooling.
- What about a new report I’ve known for 1 week? Spend the first 4 meetings on context and trust; topics emerge naturally by week 5.
- What if a 1-on-1 gets canceled? Reschedule within 48 hours, not “next week.” Repeated cancellations send a stronger signal than any 1-on-1 content could.