AI people-analytics pack

People analytics without a data team.

Paste your workforce export into the AI tool you already use. Point it at a workforce export and ask a question in plain language. It computes the answer the way a specialist would, from a canonical metrics library, and explains it so you can defend it. No data team, no dashboards, no Python.

npx skills add steven-shoemaker/peoplenometry-skills

Installs with the skills package, then run any /skill inside your AI tool.

Works with
Claude Code Cursor OpenAI Codex + any tool with skills
Open source · github.com/steven-shoemaker/peoplenometry-skills

Two ways to install

Same toolkit, whichever way you work.

No terminal

Add it in the Claude app

Download the pack, then open Settings → Skills → Upload. One file, the whole toolkit. Works in claude.ai and Claude Desktop.

Download the pack (.zip)

Settings → Skills → Upload → pick the file. That's it.

Command line

Install via the CLI

For Claude Code, Cursor, Codex, or any tool that supports skills. Installs the four skills as separate slash commands.

npx skills add steven-shoemaker/peoplenometry-skills

Then run any /skill in your tool.

Accessible on the way in. Rigorous on the way out.

Every skill runs the same loop, so you never stare at a blank cell. The analysis is done properly and shows its work, so the number survives the meeting instead of getting picked apart in it.

1Question
Not "run a report," but "why is engineering attrition up?" Start with what you're actually trying to answer.
2Data
Point a skill at your export. Messy is fine. /clean-my-export gets you to an analysis-ready file first.
3Analysis
Proper rates, fair baselines, honest caveats. Formulas come from a shared metrics library, not guesswork.
4Narrative
You get the plain answer and the methodology underneath, not a chart you still have to explain.

One library behind every number.

Sixty-plus people metrics, each defined once, the right way. The skills compute from it, so two people running the same question on the same file get the same answer. That's the line between real analysis and pasting a spreadsheet into a chatbot.

Headcount & Workforce Attrition & Retention Mobility & Progression Leadership & Org Design Performance Learning & Development Talent Acquisition Compensation Compliance & DEI Workforce Planning Employee Experience Risk & Governance

It blocks the classic mistakes

Attrition on average headcount, not terminations over everyone. Retention that isn't just one minus attrition. Never the raw pay gap. Never the average eNPS.

It's honest about its limits

Metrics that need your ATS, LMS, or survey platform say so instead of guessing. Models like flight risk are never dressed up as a single formula.

It learns your company

Definitions you confirm, and metrics the library doesn't cover yet, save locally and get reused. It gets more accurate, and more yours, over time.

Read the library →

What you'd actually say

The front door is /people-analytics. Ask it anything about your workforce data in plain language and it computes the answer. When a question needs to go deep, it hands off to a specialist.

Ask it anything

Plain language. One skill answers all of it.

"What's our headcount growth been this year, and is it hires or churn?"
"How long do people actually stay before they leave?"
"Are we top-heavy? What's our average span of control?"
"How diverse is engineering, by level?"
"What's our attrition, and is it actually a problem?"
"What's our promotion rate, and who's not moving?"

Or reach for a specialist

When you already know what you need.

Run /clean-my-export on this BambooHR file and tell me what's missing.
Run /why-are-people-leaving for the deep attrition dive.
Run /write-a-report to turn any analysis into a report.

A report, not a spreadsheet.

  • A plain-English answer up front, the story you can send up and act on.
  • The methodology underneath: formulas, sample sizes, caveats, so it survives "how did you control for tenure?"
  • One self-contained HTML file that prints straight to PDF. No dashboard, no login.
  • Rebrandable. Change a few tokens and every report carries your company's identity.
View a full example report →
report · why are people leaving?
◆ Northwind Labs · People analytics brief
People leave early: 57% are gone inside year one
Company attrition is a healthy 15.5%. The story is the timing, not the size. Most departures happen in the first year, which points at hiring and onboarding, not retention.
<6 mo
165
6–12 mo
110
1–2 yr
105
2–3 yr
80
Leavers by tenure at exit · n=486

About

The Peoplenometry method, made available to AI.

Peoplenometry is a people-analytics practice followed by a community of HR and People leaders. This pack makes that method executable inside the AI tools you already use, so an HR generalist can produce specialist-grade analysis without a data team behind them.

No data to practice on? Every skill can pull a safe, realistic synthetic workforce set from peoplesets, so you can learn the method without ever pasting a real employee record. Built by Steven Shoemaker.