How to Talk About AI Usage in 1:1s (Without Making It a Performance Metric)
2026-11-15 · 8 min read · 1:1 Meetings
AI coding tools are changing how your engineers spend their day. Some have rebuilt their whole workflow around an agent. Some use autocomplete and nothing else. Some have tried everything and quietly gone back to typing.
You should know which, and the 1:1 is the right place to find out. It's also the easiest place to get it wrong. Ask the wrong way and "how are you using AI?" sounds like "why aren't you faster?" — and you'll get a defensive answer, not an honest one.
This guide covers how to bring AI into your 1:1s so it helps the engineer, the team and you, without turning into a performance metric.
Why the 1:1 Is the Right Place
AI use is personal in a way most tooling isn't. It touches how someone thinks, how confident they feel in code they didn't type, whether they worry about their role, and how they learn. A team meeting is the right place to share workflows. A 1:1 is the right place to understand this person's experience of them.
It's also where the conversation can be two-way. Engineers often have strong views on AI tools — about quality, about what they're being pushed toward, about what's worth the licence. You want to hear those views before they show up in a resignation letter or a quietly ignored mandate.
Set the Frame First
Before the first conversation, say plainly what this is and isn't. Something like:
"I want to understand how AI tools fit into your work — what's helping, what isn't, what you'd want from me. This isn't something I'm going to use to compare people or in reviews. I'd just like to know."
Then keep that promise. If AI usage ever shows up in a performance review as a number, every future conversation about it will be guarded.
Questions That Open Conversations
Good questions are open, specific to their work, and curious rather than evaluative.
About their workflow
- "Walk me through the last thing you used an agent for. How did it go?"
- "Where has it saved you real time? Where has it cost you time?"
- "Is there a kind of task you've stopped doing by hand?"
About quality and confidence
- "How do you review code an agent wrote compared with code you wrote?"
- "Has anything slipped through that you'd have caught by typing it yourself?"
- "Do you feel you understand the code you're shipping as well as before?"
About learning and growth
- "Is AI helping you learn new areas, or skipping the learning?"
- "What would you want to get better at, with or without these tools?"
About the team
- "Is anyone on the team using these tools in a way you'd like to learn from?"
- "Would you be up for showing the team your setup?"
Questions That Shut Them Down
Avoid anything that sounds like a judgement or a quota:
- "Why is your AI usage lower than everyone else's?"
- "How much faster are you with Copilot?"
- "Leadership wants us to be AI-first — are you on board?"
- "Can you use AI more on this so we hit the date?"
Each one tells the engineer that AI usage is being measured against them, and the honest answer becomes risky to give.
Using Data Without Weaponising It
If you can see agent-assisted work in your team's Git history (see how to tell which agents your team uses), it's tempting to open with the number. Use it as a prompt for curiosity, not as evidence.
Better:
- "I noticed you've been using Claude Code a lot on the billing service lately — how's that going?"
- "It looks like most of your agent work is on infra chores. Is that where it fits best for you?"
Worse:
- "Your AI score is 12 and the team average is 40."
Two reminders whenever you look at the data:
- It's a floor. Git only shows agent work. Someone with no signatures may use autocomplete all day.
- Zero means "none detected", not "doesn't use AI" — and even when it does mean that, that may be a perfectly good choice for their work.
The Engineer Who Doesn't Use AI
Every team has someone who isn't using coding agents. Don't assume it's a problem to fix. Ask why, and listen:
- "It doesn't fit my work." Some work — debugging production incidents, design, deep refactoring in unfamiliar code — benefits less. That's a legitimate answer.
- "I tried it and the quality wasn't there." Often useful feedback for the whole team. Ask what they tried and where it fell down.
- "I don't know where to start." This is where you can help: pair them with someone who has a workflow they like, or give them a low-stakes task to try it on.
- "I'm worried about what it means for my job." Take this seriously. Talk about how the role is changing, what skills matter more now (review, design, judgement, understanding the system), and what you value in their work.
What you shouldn't do is set a usage target. Mandated adoption produces performative adoption.
The Engineer Who Uses It for Everything
The opposite needs a conversation too, gently:
- Can they still explain the code they're merging?
- Are their PRs getting larger and harder to review?
- Are they still building the understanding they'll need for the next level?
Frame it as protecting their growth, not limiting their tools: "You're shipping a lot with the agent, which is great. I want to make sure you're still getting the depth on the payments system you'll need to lead it next year."
Make It Part of the Rhythm
AI doesn't need a dedicated 1:1. A light touch works better:
- Once a month or so, add a question about tools and workflow to the agenda.
- Note what you learn alongside the rest of your 1:1 notes, so you can see how someone's use changes over a quarter.
- Follow up on offers. If someone said they'd demo their setup, make it happen.
- Bring patterns to the team, not names. "A few people have found agents great for migrations — let's share how" works. "Priya uses AI the most" doesn't.
For the broader craft of 1:1s, see the complete guide to 1:1 meetings, and for why the safety of these conversations matters so much, psychological safety in engineering teams.
How Emtricks Helps
Emtricks shows each engineer's AI adoption — their share of agent-assisted commits and PRs, how steady it is, and which tools and repos it shows up in — on their 1:1 page, right next to their GitHub activity, goals and impact stories. It's there as context for the conversation, labelled as a floor, and it never appears in a ranking. Team totals live separately in Pulse.
Key Takeaways
- The 1:1 is the right place to understand how AI fits into each person's work.
- Set the frame first: this is about understanding, not evaluation — and keep that promise.
- Ask open, specific, curious questions. Avoid anything that sounds like a quota or comparison.
- Use data as a prompt, never as evidence. It's a floor, and zero means "none detected".
- Listen to the engineer who doesn't use AI, and gently check depth with the one who uses it for everything.
Bring each engineer's GitHub activity and AI adoption into your 1:1 prep automatically. Try Emtricks free for 30 days.