Copilot

Ask Copilot a plain-language question about your team's activity, time, and money, and get an answer grounded in the same data you can already see, with every claim traceable to its evidence.

Who this is for

Anyone signed in can open Copilot. There's no special permission for it. What you can ask about is scoped to what your own role can already see, and every question is written to the access log.

What you'll see

The header reads Copilot, with the context line "Scoped to your role · every query logged."

An Ask anything box lets you type a question (placeholder: *"Ask about your team's activity, grounded in the numbers, never invented."*), with an Analysis window control set to 7 days, 30 days, or 90 days, and an Ask button (or ⌘↵) to submit it. Before you've asked anything, four suggested questions appear as buttons:

• "Who was most active this week, and on what?"
• "Which apps did the team spend the most time in?"
• "Summarize the team's focus vs. meeting time."
• "Is anyone showing signs of a heavy meeting load?"

Once you ask, an Answer card appears, tagged AI or Snapshot only depending on how it was produced:

• A grounded AI answer shows flowing prose with small citation markers after each cited sentence (hover one to see the fact it points to) and a Citations list underneath tracing every marker back to its evidence.
• If nothing could be grounded, you'll see "No grounded answer" instead: *"The copilot couldn't ground an answer to this question in the available evidence, so it declined rather than guess. The snapshot below is everything it had to work with."*
• Without natural-language generation configured, you get the plain snapshot facts instead of prose, with a note that natural-language answers need an API key configured.

Every answer is followed by an evidence table listing each member's active time, score, work/meeting/idle split, and top app for the window the answer covers.

Further down, Team digest lets you pick Today, Last 7 days, or Last 30 days and select Generate digest for a short written summary of how the whole team worked over that window: the same evidence-grounded facts as Ask, written up as one passage, with the same evidence table beneath it.

What happens next

Copilot never goes beyond what the evidence table shows: a grounded AI answer or digest is prose you can check sentence-by-sentence against the citations; a snapshot-only one shows you the same facts without commentary.

If it doesn't work

• If asking a question or generating a digest fails because you don't have access, you'll see "Copilot is unavailable": *"You don't have access to this workspace's data."*
• If a question fails for another reason, you'll see the server's own message, or "Couldn't ask the copilot."
• If a digest fails for another reason, you'll see the server's own message, or "Couldn't generate the digest."

Related

• The agent crew (overview)