Team analytics
Help your whole team get better with AI.
See what your team does well, what to coach next, and who’s growing, across ChatGPT, Claude, Gemini and Copilot. Everyone sees their own progress, and prompts stay private.
- Examples +9
- Direction +5
- Context +3
Access to AI isn’t the same as skill with it.
Seat dashboards tell you who has AI and how often they open it. They can’t tell you whether people use it well. That’s the gap Worsley measures.
Seat dashboards show
- Who has a seat
- How many chats they started
- Which tools they opened
Worsley shows
- How well each session was prompted, scored with ROIP
- Which parts of prompting the team is strong and weak at
- Whether people are getting better over time
Measured, and still trusted.
Being measured can feel like being watched. Team analytics is built so that the person being scored gets the most out of it, which is why the people who lead and run the team can trust what it shows.
- For each person
“Is my manager reading my prompts?”
No. You see your own growth. Your prompts stay yours, and the team sees scores, never what you wrote. Nobody is ranked.
- For the team lead
“I can’t coach what I can’t see.”
See what the whole team finds hard and notice who’s improving, instead of guessing.
How it works for a team
Three steps, and nobody changes how they work: people keep using the AI tools they already use.
- 01Create a team
Anyone can create a team, then invite people by their email address or share its join code.
- 02Form teams
People create or join teams, usually one per project.
- 03Capture and score
Worsley captures sessions in ChatGPT, Claude, Gemini and Copilot in the browser, and the sessions people run in their terminals. ROIP scores every one.
Know what to coach
Scores say where the team stands. These say what to do about it.
- Context is the lowest component across most of the team.
- Task decomposition lags on Gemini sessions.
- Examples rose 9 points over the last 30 days.
- Insights
- Patterns no single chart shows, written from the same deterministic scores. Rule-based: no AI writes them.
- Where you stand
- Each person sees which quarter of the team they’re in, and their own trend, never a rank.
Prompts stay private by default.
People who worry their prompts are being read switch capture off, and then the team has no data. So privacy is part of how Team analytics works, not a setting bolted on.
Scores are shared inside a team, unless the owner keeps them to owners and admins. Prompts are never shown to the team.
| Data | You | Teammates | Team admins |
|---|---|---|---|
| Your prompts | Yes | No | No |
| Your scores | Yes | Unless the owner turns it off | Yes |
| Team results | Yes | Yes | Yes |
- Prompts are private. No one on a team can read your prompts or the AI’s replies, or see your session titles.
- Your Capture Rules are yours, in your own settings.
- No one can switch capture on for you, or override you switching it off.
Roles and controls
Roles are per team: the same person can be a member of one team and an admin of another.
| Capability | Member Most people | Admin Coaches a team | Owner Runs a team |
|---|---|---|---|
| Inside a team | |||
| See the team’s results and insights | Yes | Yes | Yes |
| See teammates’ scores (when the owner allows) and where you stand | Yes | Yes | Yes |
| Invite people, manage the join code and remove members | No | Yes | Yes |
| Change roles and the team’s permissions | No | No | Yes |
| Hand over or delete the team | No | No | Yes |
Settings the owner controls
- Members See ROIP
Members see the team’s ROIP and each other’s, or it stays with owners and admins.
- Who Can Invite
Only owners and admins, or everyone on the team. Members invite as members.
- Join Code
Let people join with the team’s code, or turn it off. A new code stops the old one working.
Rolling it out
Getting everyone on the team takes an invitation or a join code.
- Invites and join codes
- Invite people by their email address and choose their role, or share the team’s join code. The owner can turn the join code off.
Questions teams ask
Can my manager read my prompts?
No. Teammates and admins see your scores and activity, never your prompts, replies or session titles.
Does Worsley use AI to grade us?
An AI model reads each prompt: Jev, a decision model from TypeSafe AI that answers fixed questions about what the prompt left open and writes nothing. A fixed formula, the same for everyone, turns those answers into the score, and team insights are rules built on those scores. Written feedback, Try Instead rewrites, session insights and the coach are worded by Claude Haiku through Anthropic’s API, only when you ask for them: it receives your prompts, and for feedback on one turn a shortened reply, and it never sets the score.
Which AI tools does it work with?
ChatGPT, Claude, Gemini and Copilot in the browser, plus Claude Code, Codex CLI, Gemini CLI, Grok Build, Qwen Code, GitHub Copilot in VS Code and local models.
Help your team get better at AI, together.
We’ll walk through the team view, the privacy model and rollout with you, and set up an early pilot if it fits.