Read schoolAgent conversations

Admin › Agent › Conversations lists the conversations people have had with your agents. Use it to see what families and staff are asking, check an answer someone complained about, or follow up on a conversation that was routed to your team.

You need Manage the agent to open it. You see the conversations in your reach: a district-level role sees its district's, and a school-level role sees its schools'.

What's listed#

The list includes schoolAgent's website chats and signed-in chats. Otto conversations aren't included. The line under each summary names the agent. Newest first.

Refresh at the top reloads the list.

Counts#

The numbers at the top count the conversations that match your filters:

• Conversations. All of them.
• Answered. The agent replied, and the conversation wasn't routed.
• Positive feedback. At least one answer was marked helpful.
• Routed. The conversation was sent to one of your routes, like a ticket or an email to a team. See Route questions to a person.
• Pending. The agent never replied, and it wasn't routed.
• Flagged. Conversations someone flagged for review (see Review).

Answered, Routed and Pending don't overlap, so together they add up to the total.

Filters#

In screen order. Role, Sentiment, Status, Rating and Flags are under Filter; each shows as a chip you can clear.

• Search. Finds words in the conversation's summary, in any message, or in the person's name or email.
• Users. One or more people. Guests can't be picked here.
• Role. The role of the person who asked, from the roles that appear in your conversations. Guest is someone who wasn't signed in.
• Sentiment. Positive, Neutral, Negative or Safety Concern. This is a simple word match on the conversation's title and the person's own messages, not the agent's replies, and not AI: "thanks" or "helpful" counts as positive; "not working" or "wrong" as negative; "suicide", "self-harm" or "unsafe" as a safety concern. A safety concern wins over the others, then positive, then negative. Only whole words count, so "badge" doesn't read as negative. Treat it as a rough guide.
• Status. Answered, Routed or Pending, as in the counts.
• Rating. Positive (at least one answer marked helpful), Negative (at least one marked not helpful) or Unrated.
• Flags. Flagged for review: conversations someone flagged from the conversation itself.

Filters apply to the counts and the list together.

Which filter should I use?#

Sentiment and Rating both look like "was this good or bad", but they measure different things:

| Filter | What it tells you | Use it when | | --- | --- | --- | | Rating: Negative | The person selected thumbs down on an answer. | You want answers people said were wrong or unhelpful. The most reliable signal, but most people never rate. | | Sentiment: Negative | The person's own words matched a word like "wrong" or "not working". | You want frustrated conversations, rated or not. Expect some that are about a broken device, not a bad answer. | | Sentiment: Safety Concern | The person's words matched a safety word. | You're checking that nothing concerning slipped by. Read each one: a word match isn't a judgment. | | Status: Pending | The agent never replied, and nothing was routed. | You're looking for chats that went nowhere. |

Examples:

• After adding a new bell schedule to the agent's knowledge, search "bell" with Rating Negative to see whether families still got a wrong time.
• A middle school tech assigned to one building opens the page and sees only that school's conversations. Status Routed with a search for "Chromebook" shows every device question the agent sent to a person there, newest first.

The list#

• Date. How long ago for the last day, then the date, with the time underneath. Uses the last activity in the conversation.
• User. Their name and email, or Guest, and the network address the chat came from, when it was recorded.
• Summary. The conversation's title, or its first message, then the number of messages and the agent's name.
• Status. Answered, Routed or Pending, plus icons for a helpful or not helpful rating, a safety concern and a flag for review. Hover over Routed for the route it went to, and over an icon for what it means.
• The ... menu at the end of a row has Open and Download as PDF. The PDF is that conversation's transcript, the same file as the download on the conversation itself: each answer's sources are listed under it, and the file name ends with the start of the conversation's ID so two chats with the same title don't overwrite each other.

Select Date, User or Summary to sort by it; select again to reverse. Sorting covers every page, not just the one you're on. Drag a column's edge to resize it, or use Columns to hide and reorder them; your browser remembers both.

20 conversations show per page; choose 50 or 100 under Rows per page. Use the arrows to move between pages.

Export CSV#

Export CSV downloads every conversation that matches your filters, not just this page, with the full transcript in one column. Other columns include the date, name, email, role, agent, sentiment, status, the route it went to, rating, whether it's flagged, network address and number of messages.

It stops at 5,000 conversations. If there are more, it tells you: narrow the filters, by role or rating for example, and export again.

A conversation#

Select a row to open the whole conversation.

The header#

The conversation's summary, and who it was: their name, role and email, or Guest and the network address for someone who wasn't signed in. Under it, badges:

• Answered, Routed to the route, or Pending, as in the list.
• Helpful or Not helpful, when the person rated an answer. Helpful wins when they rated some each way.
• Sentiment, from the word match described under Filters.
• Website chat or Signed-in chat: where it happened.
• The date and time it started, and how long it lasted, from the first message to the last.

On the right:

• Download PDF saves the transcript as a PDF, with each answer's sources listed under it.
• Create ticket starts a new ticket with the conversation filled in: the summary as the subject, the transcript as the description, and the person as the requester when they were signed in. Check it, then create it. The ticket is linked to the conversation and its source is schoolAgent. You need Work as helpdesk staff (and schoolHelpdesk) to see this button.
• Open ticket # replaces Create ticket when the conversation already became a ticket, whether it was routed to one or someone created it here.
• The arrows open the previous and next conversation on the page of the list you came from, in its order. Press K and J to do the same. They don't show when you opened the conversation from a link.

The chat#

The conversation reads as it did in the chat: the person's messages on the right, the agent's on the left, a line for each new day and the time under each message.

• When the agent offered choices, like which device, they show inside its message, and the one the person picked is highlighted.
• Sources show under the answer they were used for. Select one to open it.
• Helpful or Not helpful next to an answer shows how the person rated that answer. A rating that wasn't recorded against one answer counts in the badge and in Outcome, but shows next to no answer.
• When a conversation sat for an hour after its last message and wasn't routed, it ended: the next message started a new conversation. A note at the bottom says so.

Outcome, Person and Details#

• Outcome. How it ended, and the number of answers, helpful and not helpful ratings, and sources used.
• Person. Their name (select it to open their profile), role, email and school. Their devices are listed, each linking to the device, when you can see assets. A guest shows as Guest.
• Details. Where it happened, the agent, when it started, its last message, the number of messages, the network address when it was recorded, and the conversation's ID with a button to copy it.

Review#

• Flag for review marks the conversation so it can be found again: it counts under Flagged and shows with Flags on the list. Remove flag takes it off. Both can be undone from the confirmation.
• Add to knowledge opens the person's first question and the agent's last answer, ready to edit into a verified answer. It's saved as a draft; publish it on the Knowledge tab of Admin › Agent › Configuration. You need Manage knowledge to see it.

Routed says whether, and where, the conversation was routed.

You can't reply to or delete a conversation here. schoolOS Admin can open conversations but can't flag them.