Open the conversations behind a number
Click any number and the inbox opens filtered to exactly the conversations it counts, for the same dates, channel and topic. The filters show as removable “Analytics” chips in the inbox, and the list total matches the number on the report.Download as CSV
Each section (volume, channels, topics, handoff reasons, knowledge documents, knowledge gaps) downloads as CSV for the selected dates and filters.AI performance
AI Resolution Rate. Conversations the AI Agent handled with no human message and that the same customer did not re-open on the same topic within 72 hours, over conversations with a known outcome. Resolutions are split into Confirmed, Assumed and Re-opened. Conversations Aide is still handling count once they end or are handed off. Handle rate. A chart of ended conversations, day by day (week by week past about six weeks), split into Handled (Aide handled it without anyone on your team stepping in, and the customer didn’t re-open it) and Handed off (everything else). A conversation has ended when it was closed or the customer hasn’t replied for 72 hours. The headline shows Aide’s share and how it moved against the previous period. Involvement and quality:- Automation Rate: conversations with at least one procedure applied, split into full (Aide replied) and partial (Aide ran an action without replying).
- Topic detected: share of conversations where Aide recognised the topic with enough confidence to act on it.
- Not routed to Aide: conversations Aide was not allowed to handle: the topic is set to no automation, the channel is off, or the usage cap was reached.
- Not re-opened: Aide resolutions the customer did not re-open on the same topic within 72 hours.
- CSAT: share of survey responses rating 4 or 5 on conversations Aide handled end to end.
- Verified by your team: items your team reviewed and marked correct, across topic detection, procedures and drafts.
- Median first response, on conversations where Aide replied first.
- Team response after handoff: time from Aide’s last message to the first teammate message.
Overview
- Conversations: volume by day, split into AI-handled and Handed off.
- AI Resolution Rate, CSAT for conversations a teammate handled or took over, Median first response where a teammate replied first, and Median time to resolution (from the first customer message to resolved, including reopens).
- Sales from Aide chats.
- By channel and By topic.
Knowledge use
Documents Aide used lists the help documents Aide’s chat replies drew on, how many conversations cited each one, and how often those conversations were resolved (once a document has 30 conversations with a known outcome). Knowledge gaps groups by topic the conversations Aide handed off for missing information or answered without enough knowledge, each opening its conversations, so you know what to write next. For suggested text to add, checked against your knowledge, see AI Agent > Quality > Issues > Missing knowledge (Find and fix gaps in your knowledge).Impact
Impact puts two things side by side: five of your numbers over a period, and what Aide learned and applied during it (see What Aide learns every night). Pick a period of up to 90 days; it opens on the last 42. The five numbers. AI Resolution Rate, Acceptable replies, Drafts sent, CSAT and Median first response. Each shows its figure for the period and, after “Before”, the figure for the period of the same length just before it, with the change between the two. Acceptable replies and Drafts sent are estimates from the drafts Aide reviews each night, as on the Quality page, for all your agents together; Drafts sent counts a draft your team sent as is or with minor edits. A number with too little behind it reads “Not enough signal for this range”; it is never shown as zero. CSAT needs 30 survey responses. The trend. Select a number to draw it. AI Resolution Rate, Acceptable replies and Drafts sent are drawn as a 7-day average: each point is the rate over the 7 days up to it, because one day alone is too few conversations to read. CSAT and Median first response are drawn as a few points across the period: each point covers the days up to it, and the card says how many. Markers. A marker sits on the line on each day a learning was applied, whether Aide applied it or a teammate did. A number on a marker means several were applied that day. Select a marker to see only its learnings in the list below, and Show all to bring the rest back. A marker says when a learning was applied, nothing more. A change in the line after a marker is not proof that the learning caused it: many things move these numbers, and one learning reaches few conversations. Learnings applied. Every learning applied in the period and not undone since, newest first, with the kind of change it was, the day it was applied, and “Applied by Aide” or “Applied by a teammate”. For a learning that wrote or changed a snippet or an article, the list shows how many conversations it was given to Aide in, how those conversations ended, and the most recent ones, each opening the conversation.- “Given to Aide” means the snippet or article was among the knowledge Aide had in front of it when it wrote a reply in that conversation. It does not mean the reply quoted it.
- No learning gets a percentage. The conversations behind one learning are too few to say what it changed, so a learning shows counts and the conversations themselves. Before and after figures belong to the five numbers, over the whole period.
- “Its use can’t be traced to conversations” shows on a learning that changed your Instructions or Tone & identity, or whose document no longer exists. There is no document to follow into conversations, so there is no count.
- “Not used in a conversation yet” shows on a learning whose document Aide hasn’t been given since it was applied.