How do you use AI in one-on-ones?
Dongyoung Kim, CPO
In one-on-ones, AI is well suited to repetitive work such as summarizing for prep, recording the conversation, and tracking action items. That frees the leader to listen instead of typing during the conversation, read the context, and keep promises. But don't hand evaluation or judgments about feelings to AI, and get the team member's consent first on recording and what gets kept.
What AI should handle in one-on-ones, and what the leader should
One-on-ones often fail less because of conversation skills than because of preparation and follow-through: not remembering the last conversation, missing eye contact while taking notes, and forgetting promises before the next meeting (for more causes, see Why do one-on-ones fail?). AI cuts exactly this repetitive work.
| Stage | Good to hand to AI | The leader must do |
|---|---|---|
| Prep | Summarize past conversations and action items; draft questions | Pick the topics you must cover this time |
| During | Recording and transcription | Listen, ask back, read feelings and context |
| Notes | Organize key points, decisions, support requested, to-dos | Check the summary is right, and fix it |
| Follow-up | Track action-item status; suggest next topics | Actually do what you promised |
| Reflection | Show conversation patterns such as speaking ratio | Decide what to change next time |
The rule is simple: organizing, summarizing, and tracking go to AI; relationships, judgment, and promises stay with the leader. If AI writes sentences judging a team member's ability or attitude, they start to see one-on-ones as evidence gathering for reviews and hold back.
Using AI in one-on-ones: step by step
The prompts below are written so they work in general-purpose AI tools too. But don't put team members' names or sensitive conversation content into tools your company hasn't approved. Remove names and generalize the content, or use a tool that has passed your company's security review.
Step 1. Prep: review past conversations in three minutes
The most time-consuming part of prep is rereading past notes. Give AI the recent notes and let it organize them.
Sample prompt: “Below are the notes from my last three one-on-ones with one team member. List each of the following separately. Leave out guesses and use only what is in the notes.”
- Action items not yet finished (with owner and deadline)
- Concerns or topics that came up more than once
- What the team lead said they would do last time
The leader's job here is one thing: look at the summary and pick one or two topics you must cover this time. The full prep sequence is in How do you prepare for a one-on-one?.
Step 2. Draft questions: pick from them, don't read them out
Use AI-generated questions only as a starting point. The leader knows the team member's situation best.
Sample prompt: “A team member has mentioned twice recently that work priorities keep changing. Write five open questions that don't sound like an evaluation. Leave out questions starting with ‘why,’ and include how to ask what support they want.”
Basic questions by topic are in What questions should you ask in a one-on-one?.
Step 3. During the conversation: let AI take notes, and focus on listening
With recording and speech-to-text, you don't need to stare at a laptop and type during the conversation. Instead, keep to these three.
- Check consent before recording every time: “Is it okay if I record today again, for the notes?”
- Let the team member know in advance that you can pause the recording if something sensitive comes up.
- When something important comes up, check back in words: “Let me make sure I understood.” Even with a record, confirm in the conversation.
Step 4. Notes: AI summarizes, the leader checks
After the conversation, get the AI summary in the same format each time. A consistent format makes the next prep easier.
Sample prompt: “Organize the one-on-one transcript below under these four headings. Don't write judgments about the team member's personality or attitude; only summarize what was said.”
- Key discussion (three lines or less)
- What we decided together
- Support the team member asked for
- Action items (who, what, by when)
AI summaries can be wrong. Check that what the team member said hasn't been skewed, and fix it before sharing. For example headings, see How do you structure a one-on-one doc?.
Step 5. Follow-up: carry promises into the next one-on-one
The hard part of action items isn't setting them but bringing them back up at the next meeting. When a tool tracks their status (in progress, done) and shows them on the prep screen for the next one-on-one, the leader can open with “Let me start with what I said I'd do last time.” Criteria for setting action items are in How do you set and follow up one-on-one action items?.
Step 6. Leader reflection: review your own conversation habits
AI is useful for leaders looking back at themselves. Patterns such as who talked how much, or whether the leader jumped to answers, are hard to see until you look at the record.
Sample prompt: “Looking only at what the team lead said in this transcript, find the following three. Then suggest one thing to change next time.”
- Roughly how much the team lead talked
- Moments where they cut the team member off or gave an answer right away
- Questions that got the team member talking more
If the team member says little, see also What should you do when a team member won't talk in a one-on-one?.
General-purpose AI tools vs dedicated one-on-one tools
All of these steps can be done with a general-purpose AI tool and documents. But as the team grows, collecting and pasting notes becomes a burden of its own. At that point it's worth looking at a dedicated one-on-one tool that connects prep, recording and summaries, action-item tracking, and conversation feedback in one flow. What to look for is in How do you compare one-on-one meeting tools?.
Checklist before bringing AI into one-on-ones
AI notes are convenient, but they can weigh on team members. Before you start, settle these five things and explain them to the team first. If you work in HR, it's best to set them as company standards.
| Item | What to decide | How you might explain it |
|---|---|---|
| Purpose | What the records are for | “I use it so I don't forget my promises and can focus on our conversation.” |
| Consent | How consent is asked before recording, and whether people can decline | “If you'd rather not, we'll go without recording, and you can turn it off midway.” |
| Access | Who can see the records and summaries | “Only the two of us see the notes.” (match your actual settings) |
| Link to reviews | How one-on-one records are referenced in reviews | “They're not for rating you; they're for remembering your work and growth.” |
| Security | Where data is stored, encryption, how long audio is kept, whether it trains AI | “Audio is deleted after it's processed and isn't used to train AI.” (check the tool's actual policy) |
If it's unclear how records feed into reviews, team members get careful in front of the recorder. For the line between one-on-ones and review meetings, see How is a one-on-one different from a performance review?, and for how to explain it to the team, How do you introduce one-on-ones to your team?. Check the legal requirements for recording and personal data with your company's legal or privacy team.
Frequently Asked Questions
- Q.Is it okay to record one-on-ones?
- With the team member's consent, and with the purpose and who can see the records explained in advance, recording helps a lot. Don't treat consent as one-time; check lightly each time. It's safest to confirm company rules and legal requirements with the people responsible.
- Q.Can I summarize one-on-ones with a general-purpose AI like ChatGPT?
- First check whether your company allows the tool. If it doesn't, remove names and anything that identifies a person, and leave out sensitive topics such as health, family, or thoughts of leaving. Also check the tool's policy on whether your input is used for AI training.
- Q.Should I share the AI summary with the team member?
- Yes, it's better to. Team members trust the record when they can check that what they said was summarized correctly and fix it. Action items in particular carry into the next one-on-one only when both of you see the same list.
- Q.Can one-on-one records be used in performance reviews?
- They're useful for recalling what someone did and how they grew. But if concerns or mistakes a team member opened up about become grounds for marking them down, candid conversation disappears from one-on-ones. Decide how they'll be used before you start, and tell the team. See How do one-on-ones complement performance management?.
- Q.Can AI replace one-on-ones?
- No. AI cuts the time spent on prep and notes, but trust builds when the leader listens and keeps promises. Why one-on-ones are a leader's job is covered in Why do teams hold one-on-ones?.
ORBLIT
How Orblit helps
To make it easier to put this article into practice, Orblit helps in these ways.
01Prep to follow-up in one flow
Based on the team member's prep and past records, Orblit AI suggests questions for each team member; recorded conversations are transcribed and organized into a summary and action items, which carry into the next one-on-one's prep screen with in-progress or done status.
02Leader reflection included
AI leadership coaching shows the speaking ratio of leader and team member, and analyzes what the leader said to point out strengths and areas to improve.
03Meets security standards before you start
Meeting data is not used to train AI models, and recordings are deleted automatically within three days after analysis. Data is separated by workspace and access is managed by role.
