2 cards sit under the charts on Insights and answer the same question from opposite ends: how much of the work is being done by people, and how much of it is being done without them.
What you need: the source and Date range pickers at the top of Insights set both cards, so change the scope there rather than looking for a filter on the card.
Team productivity
One row per person with access to the sources in scope, with 4 columns.
- Member shows the name and a badge: Admin, Manager, Moderator, or Invited for somebody who has not signed in yet.
- Replies sent counts public and private replies that person sent by hand.
- Reviews handled counts every review they acted on in any way, not only the ones they replied to.
- Avg response is their own time from a review landing to their reply. It reads "Not yet" for anyone who has not replied inside the period.
The card only counts work done inside ReviewComet. A reply somebody typed in Meta's own tools is not visible to us, so a person who works half the day in Business Suite will look quieter here than they are. That is worth saying out loud before anybody reads this table as a scoreboard.
The team productivity table with a row per member showing replies sent, reviews handled and average response.
If the card is empty it says nobody on the team has access to this source. That is an access answer, not a workload one: check who is limited to which locations on the Team tab in Settings.
What the differences usually mean
- High Reviews handled, low Replies sent is someone doing the sorting work: archiving, assigning and approving. It is real work and the feed stops without it.
- A long Avg response on one person is more often an assignment problem than a speed problem. Reviews sitting assigned to somebody on holiday still count against them.
- Everyone slow on the same days is a volume problem, not a people problem. Check Review volume for the same range and look at what happened that week.
Automation and agent performance
The second card lists every rule and AI agent scoped to the sources in view. Each row carries the name, an Agent or Rule badge, and a line reading how many times it fired: "412 triggers, 388 replies sent", or "412 triggers, no replies" for a rule that flags, assigns or notifies rather than answering.
The small chart on the right of each row is the trend across the period. It is there to show shape, not exact figures: a line climbing at the end means that rule is picking up work, and a line that falls off a cliff usually means the thing it matched stopped happening or something upstream started catching it first.
The automation and agent performance list with a trend sparkline on each row.
If no rule or agent is scoped to the selected source, the card says so instead of showing an empty list.
What to do with the differences
- A rule with a lot of triggers and no replies is flagging, assigning or notifying. Check it is catching what you meant it to catch before you leave it running.
- A rule that never fires is either too narrow or has been overtaken by one above it. Backtest it rather than guessing.
- An agent replying far more than the rules is usually the sign that a fixed reply would cover most of it. Rules are cheaper to reason about, and a saved reply is free to send by hand.
- Everything firing more, response time flat means automation is absorbing the growth. That is the number worth reporting.
Every trigger on this card is an automated action: a rule trigger spends 1 credit, and an agent reply costs more than a check. Manual work by the team costs nothing, however much of it the first card shows.