Chat Logs is where you see exactly what your agent has been saying to visitors. Every conversation it has, whether it's one exchange or a long back-and-forth, is recorded here in full and lined up in a simple list so you can scan through them.
It's also the fastest way to fix a wrong answer. Instead of guessing what your agent might be getting wrong, you read the actual conversation, mark the bad reply with a thumbs-down, and correct it on the spot. That correction is saved as a Q&A pair immediately, so the next visitor who asks the same thing gets the right answer.
The conversation list
What it is
Chat Logs opens to a list of every conversation your agent has had. Each row shows the visitor's Name and Email (when available), the Last human message and Last agent message, and the time of the Last activity. When conversations came from eWebinar, a Webinar column also shows which webinar each one happened on, since one agent can be attached to several webinars at once.
What it's best for
- Getting a quick read on how much your agent is being used, and by whom
- Seeing at a glance which conversations ended without resolution, based on the last messages shown
- For eWebinar agents, seeing which specific webinar each conversation came from, which matters when one agent covers several session times
How it works
Open Chat Logs from the Analytics section of your agent. The list is sorted by most recent activity first, so new conversations always surface at the top. Click any row to open the full conversation; while a conversation is open, the list narrows and hides the two message columns to make room.
When the AI is notified about a notify-moderators interaction card and stays silent, no AI reply is written for that turn — that is expected. The interaction answer still appears in the eWebinar chat as before. Credits for that look still appear in your usage ledger.
Replaying a conversation
What it is
Opening a conversation from the list replays it exactly as the visitor experienced it. This isn't a summary or a simplified transcript - it's the real chat interface, in view-only mode, rendered message by message in the order the visitor saw them.
What it's best for
- Judging tone and accuracy in context, not just reading an isolated answer
- Seeing whether an escalation prompt appeared, and whether the visitor accepted or declined it
- Understanding a confusing exchange by seeing exactly what the visitor asked first
How it works
Because the replay uses the same chat interface visitors actually use, what you see is what they saw - message bubbles, timing, and order included. There's nothing to interpret or translate; you're looking at the real thing.
Giving feedback with thumbs up and thumbs down
What it is
Inside a replayed conversation, your agent's replies can be rated with thumbs-up and thumbs-down. The controls show on the latest reply you can rate (never the greeting); hover over an earlier reply to rate it too. This is how you tell the agent, reply by reply, whether it got the answer right.
What it's best for
- Fixing a specific wrong answer the moment you spot it, without digging through Q&A to find where it came from
- Keeping track of which answers are already working well, not just the ones that need fixing
- Catching answers that are technically correct but off in tone or level of detail
How it works
Clicking thumbs-up simply records that the reply was good. Clicking thumbs-down opens the Add Q&A pair form with the visitor's question and an AI-drafted answer you can edit. Clicking Add Q&A saves it straight into your Q&A - you don't need to leave Chat Logs, open the Training tab, or write the pair from scratch. Because Q&A pairs take priority over every other training source, your correction takes effect immediately for future visitors asking the same question.
Best practices
Skim Chat Logs regularly in the first few weeks after launching a new agent - this is when it's most likely to say something off, and when a few minutes of thumbs-down corrections pays off the most. Thumbs-down anything wrong as soon as you see it rather than making a mental note; the correction takes thirty seconds and your Q&A quality compounds fast once you get in the habit.
Opening from eWebinar
What it is
When you spot a weak AI reply in the eWebinar chat inbox, Improve answer in eChat jumps you straight into Chat Logs for that conversation so you can fix it without hunting for the row yourself.
What it's best for
- Correcting an answer you noticed while moderating in eWebinar, without switching products and searching the list
- Landing on the exact reply that needs a thumbs-down, with the Add Q&A form already open
How it works
On an answer the agent gave, choose Improve answer in eChat in the eWebinar inbox. eChat opens Analytics → Chat Logs for that agent, finds the matching conversation, scrolls to the reply, outlines it, and presses thumbs-down for you — the same Add Q&A form you'd get by clicking thumbs-down yourself in the replay.
If you aren't signed in, you'll sign in first and return to the same conversation; the link is not dropped. After the reply is focused once, a refresh opens the conversation on the latest message as usual.
If that conversation isn't in this agent's Chat Logs (for example it belongs to a different agent, or the reply never landed in eChat), you'll see a short notice and stay on the Chat Logs list.
Best practices
Use Improve answer in eChat as soon as you see a bad reply in the inbox — it's the shortest path from "this answer is wrong" to a saved Q&A correction. For the full thumbs-down workflow once you're in Chat Logs, see the section above.
Filtering by date range
What it is
Chat Logs includes a date-range picker so you can narrow the conversation list to a specific window of time instead of scrolling through everything.
How it works
Choose a preset - Today, Yesterday, Last 7 days, Last 30 days, This month, Last month, Last 90 days, or All time - or pick Custom range and select start and end dates on the calendar.
What it's best for
- Spot-checking how your agent performed right after a big webinar, when a spike in traffic can also mean a spike in unfamiliar questions
- Reviewing conversations from just after a training change, so you can see whether the new material actually improved answers
Best practices
After a webinar with unusually high attendance, filter Chat Logs to that day and read through a handful of conversations - it's the quickest way to catch questions your training material doesn't cover yet. After you add or edit anything in Q&A or Web Pages, filter to the following day and check that the change actually shows up in how the agent responds, rather than assuming it did.