Item status reporting
At the top of the Items list, a row of status cards summarizes your items by status. Each card shows a status and a count, giving you a quick read on how much work is in progress, waiting on review, complete, or failed — without opening any single item.
Permissions. The Items list, and these cards, are available to users who can view items — the Reviewer role (its main workspace) and Admins. See Users and access.
The four primary status cards
Most items move through four lifecycle states, and these are the cards you'll see most often:
- In Progress — the item is being processed by a flow's nodes.
- Review — the item is paused and waiting for a person, having been sent to human review. See Human review.
- Complete — the item finished its flow successfully and reached the terminal state.
- Failed — the item stopped on an error before completing.
A card's count is the number of items currently in that status, so the four together tell you how the AI Agent's work is distributed right now.
What the counts represent and when they update
- Counts are for the current AI Agent, and they reflect the whole filtered set of items, not just the page you're viewing — so a count can be larger than the number of rows on screen.
- Counts follow your filters. When you filter the list — by status, reviewer, flow, date, or search — the cards recount to match the same filtered set, so the cards and the list below always agree.
- Counts refresh when the list loads. They reflect the data as of the last load; reload the list to bring them up to date. The cards are a point-in-time count, not a live ticker, and they don't show trends or rates.
A card appears for each status that has items
The cards are driven by your data: a card is shown for each status that currently has at least one item. Alongside the four lifecycle states above, you may therefore also see cards for other states when items are in them:
- Queued — accepted and waiting to start processing.
- Deferred — set aside to be picked up later.
- Closed — ended without continuing through the normal completion path.
Items in a Canceled state are not shown as a card. Status labels are written in plain title case (for example, In Progress, Review).
Drilling into the items behind a count
The cards summarize counts; to work with the items themselves, use the list below them:
- Filter by status to narrow the list to one state — for example, Review to see everything waiting on a person, or Failed to see everything that errored. The status filter offers every state, including ones not shown as a card.
- Open an item from the filtered list to see its detail, including the Insights overview and the Review tab. See Human review.
- Investigate a failed item by opening it and reviewing the run that produced it. The error detail for a run lives in Run history, which an Admin or Partner can open to read what went wrong, step by step.
How these cards relate to item statuses in the API
The statuses shown on these cards are the same item statuses the platform tracks for each item, including the ones reported through the API. The API exposes the full set of item statuses; the cards here are simply a count summary of those statuses for the current AI Agent and filters. They are a quick operational read — not an analytics dashboard, and they don't provide throughput, straight-through rates, or historical trends.
Where to go next
- Human review — act on items in the Review status.
- Run history — read the error behind a Failed item (Admin or Partner).
- Monitor and Review — how item status fits the production operating model.
- Users and access — who can view items and reviews.