Here are quick answers to the questions we hear most about how eChat works day to day. For the full picture on any of these, follow the links to the dedicated articles.
Why did my agent stop answering?
Most likely your account has run out of credits for the billing period. When that happens, eChat can't generate any more replies, so instead of leaving visitors without a response, the attendee's newest message is flagged Needs reply in eWebinar for your human moderators to pick up. Add credits, turn on auto-recharge, or upgrade your plan to bring the agent back — check credits and billing for how usage is calculated. If it's just slow to answer, or skipped one message after a moderator jumped in, see the next question.
Why didn't the AI answer right away?
Usually because it was giving your human moderators the first word. Each webinar's AI moderator settings in eWebinar have AI answers if a human doesn't answer in … sec (0 to 900 seconds; 2 minutes by default for Human hand-off agents). While a moderator is available, they're notified of the question and the AI waits that long before answering. If a moderator replies or starts typing during the wait, the AI doesn't answer that message at all. The AI skips the wait when it's already the one talking with the attendee, or when no moderator is available — and for a while after a moderator's own reply it waits at least a minute before answering the attendee's next message, so it doesn't talk over them. No credits are used while it waits. Set the wait to 0 for sessions nobody watches. See the wait time.
Can one agent handle more than one webinar?
Yes. You can link a single agent to as many eWebinar webinars as you like, and every message tells the agent exactly which webinar the visitor is on, so it automatically pulls that webinar's own video transcript and context for the reply. The one rule is to keep an agent scoped to one product, even if that product runs on several webinars — different session times, or separate sales and onboarding versions. If you sell more than one product, give each one its own agent so its training stays focused.
Does my agent's training update automatically?
It depends on the source: Web Pages re-crawl on whatever schedule you set — weekly, every 2 weeks, or monthly (or Once, if you'd rather recrawl by hand) — and Video Transcript stays synced to your webinar on its own. Conversations aren't training data, but for eWebinar agents they keep producing new suggested Q&A pairs after the first import: newly archived chats that a human on your team helped answer are queued and imported in batches (weekly, or once 50 are waiting), and nothing becomes training until you approve the resulting Q&A. Q&A and Documents only change when you or your team add, approve, or edit them yourself. See training your agent for details on each source.
Will visitors know they're talking to an AI?
That's your call. The Show as AI setting on the agent's Agent screen, alongside its name and avatar, adds an "(AI)" badge after the agent's name in the chat — or leave it off so the agent blends in alongside your human moderators. Either way the agent behaves exactly the same — this setting only changes how it's labeled in the chat window, not what it does.
What happens if the agent doesn't know the answer?
It doesn't guess. Grounding strictness controls how aggressively the agent double-checks its claims against your training material before replying, but even on Strict it never falls back to a canned refusal — it tells the visitor honestly, in its own words, that it isn't sure. Because it works alongside your moderators, it can also make sure a person sees the question, depending on your escalation mode. With Leave a note or Human hand-off, it can flag the message Needs reply and leave your moderators a private note while it keeps helping — that's not a hand-off. Only with Human hand-off (and Answer is uncertain turned on) can it offer to bring a person into the conversation, and only after the attendee agrees and if eWebinar allows it on that webinar. With Chatbot only, it never mentions a human at all.
Can I edit a bad answer?
Yes, two ways. Thumbs-down a reply in Chat Logs and save the correct answer in the Add Q&A pair form — it's added to your Q&A right away — or go straight to Training's Q&A tab and add or fix a pair yourself. Because Q&A pairs are the most authoritative training source, a correction there takes priority over whatever the agent originally learned from Web Pages, Video Transcript, or Documents.
The AI only answered my last question
On eWebinar chats, it doesn't anymore. When attendees send several messages before a moderator or the AI replies, the agent now answers the full unanswered burst in one turn — every message since the last moderator or AI reply — rather than only the most recent one. See how the AI answers for more on burst replies and how poll, quiz, and question-card answers are handled.
Does the AI reply when someone answers a poll?
When the poll, quiz, or question card has notify moderators on, yes — the AI moderator is notified the same way your human team is, and it may reply while the attendee is still watching. It can also stay silent when speaking would not help. Typed chat questions still take priority over a card. Silence still uses credits because the model ran. See how the AI answers.
Is there a live typed-out response, like the agent is typing?
No. eChat's replies arrive as one complete message once the agent has finished generating it — the text isn't streamed out word by word. In eWebinar, a typing indicator can show while the agent is writing, but never while it's waiting for a human to answer first.
Can eChat post to Slack or my CRM?
Not today. The agent involves your team only through eWebinar itself — Needs reply flags, private notes, and (with Human hand-off) hand-offs to your eWebinar moderators — not through connections to outside tools. There's no Slack, Zapier, or CRM integration to post to yet. (Your eWebinar moderator notifications, which can include Slack, are configured in eWebinar.)