Interactive answers in chats
The cards, charts, questions and agent proposals an agent can put in a reply, which ones you can act on, and where they end up.
An agent can put more than text in a reply. Alongside its words it can render cards, charts, sets of questions and proposals for agent runs, and you can answer some of them by tapping. They appear in chats, in Ask Auron and in a meeting's Chat tab.
What an agent can show you
| What appears | What it is | Can you act on it |
|---|---|---|
| Stat cards | Labelled numbers, each with an optional up, down or flat trend | No |
| Bar chart | A bar chart with a legend | No |
| Line chart | A line chart with a legend | No |
| Table | A table with defined columns | No |
| Expandable list | Items you tap to open, sometimes with one already open | No |
| Buttons | Buttons that each send a follow-up message for you | Yes |
| Single choice | A question with options: pick one and tap Confirm | Yes |
| Multi select | Several options and a Submit button | Yes |
| Question sequence | A short run of questions, each free text or a choice, sent together with Submit All | Yes |
| Agent proposal | A run an agent offers to do, with Approve and Dismiss | Yes |
Read-only cards sit inside a titled card. The ones you can act on are drawn without that frame, so they read as part of the conversation. Once you answer a choice, a multi select or a question sequence, its button changes to Submitted.
Only the latest reply is live
Buttons, choices, multi selects and question sequences work only on the most recent agent reply. On older replies they are inert, so scrolling back through a long chat cannot re-send an answer from three turns ago.
Agent proposals are the exception. A proposal stays live wherever it is in the thread, until you act on it or it expires.
Agent proposals
When you ask for work one of your agents can do, for example "Research a prospect" in Ask Auron, the reply can carry a proposal instead of the work itself. The card names the agent, quotes what you asked, and lists what the run would cover. If you asked for the work to repeat, the card shows the schedule and the next run.
Tap Approve to start the run or Dismiss to drop it. The card then follows the run:
| The card shows | What happened |
|---|---|
| Working on the subject | The run is in progress |
| It needs your input | The agent has a question for you, posted in your feed |
| Ready, with Open | The run finished and wrote something; Open shows it |
| The agent's reason, or "It finished with nothing to add" | The run found nothing worth writing |
| Stopped | The run was cancelled |
| A red error line | The run stopped before finishing |
| Dismissed | You dismissed the proposal |
| This proposal expired | A week passed without an answer; ask again to run it |
More on what agents write and where their cards land is in Agents and their cards.
They stay with the message
Cards arrive as the reply is written, and they are saved with the message. Closing a chat and coming back renders them again without asking the agent to redo the work.
The Artifacts tab
The read-only cards from the whole thread are also gathered in a chat's Artifacts tab, next to the reports. Each is labelled with its own title or question, or with its kind (Key metrics, Bar chart, Line chart, Data table, Details) when it has none. Buttons, choices, multi selects, question sequences and proposals are left out, since they are part of the conversation.