This page is the complete inventory of what AI actually does in AB Projects today. The rule behind all of it: AI drafts, summarizes, and answers — it never changes your project without you confirming, and anything AI-generated is clearly marked as such.
Everything AI does, at a glance
| Capability | Where | Can it change data? |
|---|---|---|
| Dashboard greeting | Home (personal and project view) | No — read-only orientation |
| Task summary | Task page, speech bubble at the top | No — read-only orientation |
| Suggested subtasks | Task action menu → AI: Suggest Subtasks | Only after you press Create Selected Subtasks |
| Meeting-task extraction | Extract tasks from this meeting, after a scheduled meeting ends | Only the candidates you select and create |
| Teams bot | Personal chat, or @mention in the channel | Answers questions; changes go through normal task actions |
| “Today's focus” note | Top of your daily digest email | No — commentary on your task list |
| MCP AI assistants | Claude, ChatGPT, etc., connected via MCP | Yes, on your instruction — and always disclosed in-app |
On your dashboard and tasks
- Dashboard greeting. Your Home view opens with a short AI-written summary in a robot speech bubble — a context-aware read of where things stand as you arrive, not an action plan.
- Task summary. Opening a task shows the same kind of bubble for that task: status, dates, and recent activity distilled so you don't re-read the whole history to get oriented.
- Suggested subtasks. From the task's action menu, choose AI: Suggest Subtasks. The AI Subtask Suggestions panel proposes a breakdown with editable titles, descriptions, and estimated hours; use Select all / Deselect all, press Regenerate for another take, and only Create Selected Subtasks makes them real (see Subtasks and Dependencies).
Turning meetings into tasks
AB Projects can schedule a meeting from a task (Schedule Meeting in the task's action menu: it checks participants' calendars with Find Available Slots, posts a voting card to the channel with Send Poll to Teams, and creates the Outlook invite with a Teams link once the organizer confirms). The AI part comes afterwards:
- Hold the meeting as a Teams online meeting scheduled through that flow.
- Press Extract tasks from this meeting — the button appears on the meeting entry once the meeting has ended. AI reads the meeting's transcript and proposes task candidates.
- Review, edit, and select the candidates you want, then create them. Nothing enters the project until you do.
The AB Projects bot in Teams
Talk to the bot in a personal chat, or @mention it in a linked channel. Fixed phrases get instant answers; anything else goes through AI:
| Where | Fixed commands |
|---|---|
| Personal chat | my tasks · tasks today · help |
| Channel (@mention) | team tasks · team members · tasks today · help |
- Ask in your own words. The bot uses AI to map free-form messages to the right action — “what's overdue?” or “top 10 tasks by deadline” work fine — and if the message isn't a known request, it answers conversationally.
- team members returns a card listing each member with their active task count, sorted by load — it shows the numbers; it doesn't recommend reassignments.
- The bot answers questions; it doesn't silently change the project. Edits still go through the normal task actions, with their reason and history.
Your digest and external assistants
- “Today's focus.” The daily digest email opens with a short AI note pointing at what deserves attention first — the digest itself is covered in Daily Digest and AI Assistance.
- MCP assistants. Through MCP (an open standard that lets AI tools connect to apps), an assistant like Claude or ChatGPT can create, update, query, and comment on tasks and documents on your behalf — see Connecting AI Assistants via MCP and the MCP Tools Reference. Everything an assistant creates is disclosed in-app, so it's always obvious what came from AI versus a person.
From a quick AI read of your dashboard to a bot you can just ask, tasks extracted from your meetings, and assistants that work over MCP — AB Projects keeps AI useful, visible, and under your control.