AI Profiling

The more the crew knows about your work, the less you have to explain. AI Profiling is how it learns — quietly building a picture of what you do and who you do it with, from the tools you've already connected. Everything it learns sits in one place, Settings → AI Profiling, where you can read it, correct it, or delete it.

What the crew builds

Two things:

• About you — a list of facts about your own work: your role and company, the work streams you're focused on, your work rhythm, how you communicate, and any context you add yourself. Each fact shows where it came from.
• People the crew knows — a profile for each person you actually work with: their role and company, a one-line summary of your relationship, open threads between you, communication style (yours and theirs), and related email addresses.

These aren't trivia — they're working context. When a crew task writes your meeting briefing, the attendee profiles come from here. When Captain Fokus drafts a reply or answers a question, it reads your work profile first.

Where it learns from

Profiling only reads sources you've connected:

• Calendar — meeting patterns. Anyone you've met with at least twice becomes a contact profile with your meeting cadence.
• Synced tools — your task activity across connected project tools becomes work-stream facts ("actively working on X").
• Email — your sent and received mail builds contact profiles, with a deeper pass on the ~20 people you exchange the most mail with to infer roles, relationships, open threads, and communication style.

Scans run when you connect a new tool, nightly while you sleep, and whenever you press Rescan on the profiling page. When a scan finds something new, you'll get a notification — "Your crew learned something new" — that takes you straight to the review page.

Reviewing what it learned

Facts are assumed true and put to work right away — review is how you correct course, not a gate you have to pass. On the profiling page you can:

• Edit any fact inline. Editing a machine-learned fact marks it Confirmed — your words always outrank what was mined.
• Remove a fact. Removed machine-learned facts never come back — future scans remember the rejection.
• Add your own facts ("Add something the crew should know…") for yourself or any contact.
• Delete an entire contact profile. It can only reappear if you keep interacting with that person.

Privacy and control

• Nothing is hidden. Every stored fact — yours and your contacts' — is visible on the profiling page, with the sources it came from.
• You have the last word. Edit, remove, or add facts at any time; removals stick.
• Only real relationships. Contact profiles are built only for people you genuinely interact with — a couple of meetings or a real mail thread, not every address that ever CC'd you. Personal email domains are never expanded into company-wide guesses.
• It stays in your workspace. Profiles personalize your own briefings, drafts, and answers — they're not shared.

There's no master switch for profiling; control is per-fact and per-profile. If you want a lighter footprint, remove what you don't want kept — it won't return.