Run Better 1-on-1s With Employees Using AI

Start by creating a new project in the Claude desktop app for this workflow. Claude and Cowork now live in the same app, so the project is the workspace and Cowork is the mode you turn on when you want Claude to work across the project files.

Keep the project narrow. You can create one project for your whole team, or one project per employee or mentee.

For a first pass, a team-level project is usually enough. Name it something obvious, such as:

1-on-1 Coaching System

Inside that project, add the transcripts or notes from previous 1-on-1s. In the video guide, we use a physical project folder with a few Markdown transcripts from previous calls.

Use simple filenames so Claude can tell what each file is:

2026-07-01-maya-1on1.md
2026-07-02-devon-1on1.md
2026-07-03-rina-1on1.md

Each transcript should include the date, the person, and the rough topic of the conversation. If you do not have a transcript, use meeting notes instead. A short summary is better than nothing as long as it captures what was discussed, what was promised, and what needs follow-up.

If you do not already have transcripts in a folder, use Claude's connector directory to pull them from your note-taking tool. In the video guide, we call out Granola, Fireflies, and Otter as examples of tools that can hold past call transcripts.

Before asking Claude to build the system, select Cowork in the message box and confirm it can see the files:

Before we start, confirm which transcript or note files you can read in this project.

Only list the files you can actually access. Do not summarize them yet.

This quick check prevents a common failure mode: Claude writes a generic template because it cannot actually see the source material.

Before creating the template and rubric, give Claude a compact reference file for the team. This makes the rest of the workflow less generic because Claude can see who is on the team, what they do, and what kind of calls you are running.

Use this prompt:

Take a look at the call transcripts I have in this project.

These are from my 1-on-1s with [team or employee type].

Use the transcripts to create a quick team.md file that describes:
- who is on the team
- what kind of work they do
- what the 1-on-1s are usually about
- any recurring themes Claude should remember for future calls

Important:
- Treat this as coaching support, not an HR performance review.
- Ground the summary in the transcripts.
- Do not infer private motives, health, personality traits, or personal circumstances.

In the video guide, Claude creates a team.md file for a fictional AI research team. That file becomes the project reference point for the next prompts.

If the team changes later, you can ask Claude to update team.md instead of rebuilding the whole project.

Once Claude understands the team, ask it to create the two core files: a standard call template and a standard evaluation rubric. The template helps you run the live call. The rubric helps you review each call consistently afterward.

Use this prompt:

Given what you know about the team and the information in these transcripts, create a 1-on-1 call template that I can use for monthly progress check-ins with my team.

Then create an evaluation rubric that I can use to grade my team members after the call or during the call.

Make both the rubric and the template custom to my team and how the transcripts go, but make sure they are two files that are standardized across my team.

When you are done, tell me:
- what files you reviewed
- how you used those files to create the documents
- which decisions came from the transcripts or team.md

Ground every decision in the transcripts or the team.md file.

Replace the team description and cadence with your own context. For an AI research team, the template might include research focus, proof of work, demo readiness, and next artifacts. For a customer success team, it might include account risk, follow-up quality, escalation blockers, and renewal readiness.

Review both files before you use them. A good result should feel specific to your team. It should mention patterns from the transcripts, not just generic management advice.

If the output is too vague, use this recovery prompt:

This is too generic. Revise it so every major point includes:
- the specific transcript it came from
- the observed behavior
- a short quote or paraphrased moment
- why it matters for future 1-on-1s
- one coaching question I can ask

Remove any point that is not supported by the transcripts.

If Claude returns the template and rubric as chat text instead of saved files, ask it to create the files directly:

Create two saved project files from that output:

1. one-on-one-template.md
2. evaluation-rubric.md

Keep both practical, short enough to reuse, and specific to the transcript patterns you saw.

Use the template to structure the next live call. Open it before the conversation and use it as a checklist for what to cover: progress, blockers, commitments, coaching questions, and next actions.

After the call, you have two options. You can fill out the rubric yourself, or you can ask Claude to create a first pass from the transcript.

Use this prompt:

Take a look at [Employee]'s transcript from [Date].

Create a new rubric for that call and fill it out objectively based on that transcript.

Use evaluation-rubric.md as the structure.

For each category:
- give the score
- cite the transcript evidence
- say when a category was not discussed enough to score
- avoid HR, compensation, promotion, or disciplinary recommendations

In the video guide, Claude fills out Maya's July 1 rubric and marks categories that were not evidenced by the transcript. That is the right behavior. It is better for Claude to say "not discussed" than to invent a score.

Your judgment should stay in the loop. Treat Claude's completed rubric as a draft you review, edit, and approve.

Once Claude has at least one completed rubric, ask it to create a tracker. In the video guide, we ask for a CSV that tracks 1-on-1s, team member scores, and performance over time.

Use this prompt:

Make a CSV to track 1-on-1s, team member scores, and performance over time.

Include:
- employee or mentee name
- date of each 1-on-1
- rubric category scores
- average score
- short evidence notes
- commitments made
- follow-up items

That tracker becomes the memory layer for future check-ins. After each new 1-on-1, drop in the transcript and ask Claude to update the rubric and tracker.

You can use the same project to prep future calls. Before your next 1-on-1, ask:

Look at my most recent 1-on-1 transcript and scoring rubric for [Employee].

Use them to create a short agenda for our next call.

Include:
- what to follow up on
- what improved
- what still needs evidence
- questions I should ask
- commitments to revisit

This is the loop: use the template, run the call, record or summarize the conversation, fill the rubric, update the tracker, and prep the next call from the same history.