What Is an AI Wingman?
An AI Wingman is your personal AI teammate. It learns how you work, understands your company, and does real work for you in conversation. When the work depends on a colleague, it asks that colleague's Wingman directly, so nothing waits on a meeting.
Everyone is AI-productive now. The team still hits a wall, because work depends on other people: the answer you need sits with a colleague, the context lives in someone's head, and the meeting to get it is tomorrow. Catalyst gives every team member a Wingman, and the Wingmen collaborate. That is the difference between a faster individual and a faster team.
Your day starts with it
The first time you meet your Wingman it does not ask you to set anything up. It greets you by name and brings your day:
Good morning, Emma. I'm your Wingman. Here's your day:
🔥 Priority- 📉 Annual-plan conversion has been down 31% since last Thursday's release. Want to look at that first?
- 🗂️ Your growth review deck is drafted. Take one look and I'll send it as the pre-read.
- ✅ Nina needs a yes on her campaign copy today.
Anything that needs your attention becomes a ToDo, in the Wingman's own words. Some of it is already done and only needs your look. The rest is what it will work on with you.
What makes a Wingman different
- It learns from you. It watches how you work, from the apps you use to your Claude Code and Codex sessions, your conversations and your docs, and keeps what you approve. Over time it knows your definitions, your people and your standards.
- It understands your company. It knows who owns what through the Company's Agent Directory, and it holds your own expertise so colleagues' Wingmen can ask for it.
- It does real work for you. It queries the data, drafts the deck, builds the dashboard, files the ToDos and shares the result where the team lives.
- It collaborates Wingman to Wingman. When the work depends on a colleague, it asks that colleague's Wingman in plain English and gets an answer in seconds, stamped with whose expertise it used. The colleague never looks up.
- It hands off what repeats. When a job turns into a routine, the Wingman spawns an AI Employee that runs it, watches it and reports back.
AI Wingman vs chatbot vs copilot vs autonomous agent
| Chatbot | Copilot | Autonomous agent | AI Wingman | |
|---|---|---|---|---|
| Works with | Anyone, statelessly | You, inside one app | Nobody, it works alone | You, and it learns how you work |
| Delivers | An answer | A suggestion | An unreviewed result | Finished work: dashboards, drafts, ToDos, outcomes |
| Company knowledge | None | Whatever is on screen | Improvised from documents | Colleagues' approved expertise, stamped with their name |
| Reaching a colleague | You do it | You do it | It guesses instead | Asks the colleague's Wingman directly |
| When work repeats | You ask again | You do it again | It runs unsupervised | Spawns an AI Employee that watches and reports |
A real solve: the annual-plan drop
Emma, Head of Growth, picks the priority: "Let's start with the annual-plan drop."
- Her Wingman opens the Company's Agent Directory and asks who owns annual plans. One path lights: Product → Billing & Discounts → Alex's Wingman
- It asks Alex's Wingman, in plain English: "Did anything change for annual plans since Thursday?"
- Nine seconds later: "Yes. Thursday's release changed the discount rules. The 30% code now applies to monthly plans only, so annual buyers see full price and leave." Stamped Expertise · from Alex's Wingman · High confidence.
- Then: "A fix is already in progress. It goes live in about 15 minutes." Alex never looked up from the fix.
- Emma: "Let's build a dashboard to monitor it and share it in #product." Built, live at its own URL, shared in the channel.
One conversation, zero meetings. The full loop, step by step, is in How Catalyst works; more patterns like it in Use cases.
Why the grounding matters
A Wingman never improvises company-specific work. When nothing proven covers a question, it asks a person. It does not guess.
Every company answer arrives with two things: whose expertise it used, "from Alex's Wingman", and how confident it is:
- High confidenceThe answer follows a colleague's approved expertise as written.
- Moderate confidenceThe Wingman had to make one judgment of its own, and it names it.
- Low confidenceNothing proven covered the question. The answer was worked out from the data itself.
You always know what you are reading. The full argument is in Trusted AI agents.
Where your Wingman works
- In conversation. The chat where the work happens: queries run, dashboards get built, ToDos get filed, colleagues' Wingmen answer.
- In Slack. Tag the Company Brain in any thread and get the number, the source and whose Wingman it came from.
- In your tools, in your cloud. Catalyst runs as a plugin inside Claude Code and Codex on the AI subscription you already have. The workspace lives in your own cloud and your data stays there.
Frequently asked questions
What is an AI Wingman in one sentence?
Your personal AI teammate: it learns how you work, understands your company, does real work for you, and asks your colleagues' Wingmen when the work depends on them.
How is it different from a copilot or a chatbot?
A chatbot answers and a copilot autocompletes. A Wingman brings you your day, works a problem end to end and reaches the colleague's Wingman when it needs to, so the work never waits on a meeting.
How is it different from an autonomous agent?
An autonomous agent works alone and improvises the company-specific parts. A Wingman works with one named person, uses expertise colleagues approved, stamps every answer with whose expertise it used and how confident it is, and asks a person when nothing proven covers it.
How do Wingmen collaborate?
Through the Company's Agent Directory, which maps who owns what. Your Wingman finds the owner and asks their Wingman directly, in plain English. The answer comes back in seconds with the owner's name on it, and the owner is never interrupted.
Can an AI Wingman be trusted with real work?
Yes, because trust is structural: approved expertise only, every answer stamped with its source and a confidence grade, a person approving what the Wingman keeps, and repeatable work handed to an AI Employee only after the approach has proven itself.
Meet your Wingman
Bring one real problem. Watch your Wingman find who owns it, ask their Wingman, and do the work.