> For the complete documentation index, see [llms.txt](https://help.gleantap.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.gleantap.com/ai-chatbot/managing-ai-conversations.md). # Managing AI Conversations Once your AI Agent is live, it will hold conversations across whichever channels you enabled — website, SMS, email, Facebook, Instagram, Google Business Messages, WhatsApp, and Voice. This page covers what happens *after* those conversations: where leads land, where transcripts live, how handoff to your team works, and how to fix it when notifications don't fire. For the initial setup, see Configure AI Chatbot. *** ## One agent, eight channels Your AI Agent is a **single system** operating across all enabled channels. The instructions, training data, custom intents, and tools you configured once apply everywhere it's deployed. The only per-channel setting is whether the agent is turned on or off for that channel. **Supported channels:** | Channel | Where the AI shows up | | ---------------------------- | -------------------------------------------------- | | **Web** | The chat widget embedded on your website | | **SMS** | Inbound texts to your Gleantap SMS number | | **Email** | Inbound emails to your configured address | | **Facebook** | Facebook Messenger conversations | | **Instagram** | Instagram DMs | | **Google Business Messages** | Messages sent through your Google Business Profile | | **WhatsApp** | WhatsApp Business messages | | **Voice** | Inbound calls to your Voice-enabled number | Enable channels under **Configure AI → Deploy AI to Channels**. Disabling a channel turns the agent off for that surface only; other channels stay live. *** ## Where AI-collected leads land When the AI captures a lead over a conversation (name, email, phone — configured under **lead attributes**), Gleantap does three things: 1. **Creates or updates a contact** in **Audience → Contacts**. If a contact with a matching email or phone already exists, their record is updated; otherwise a new contact is created. 2. **Writes the conversation to the Inbox** as a normal thread. You can review, reply, and reassign it like any other conversation. 3. **Logs the AI activity** on the contact's profile under the **Activities** tab — you'll see when the AI engaged with them and what happened. You can find all AI-touched contacts in one place: build a segment filtering by the source or tag your AI Agent sets (see Configure AI Chatbot for how to configure lead attributes and any custom tagging). *** ## Where to find AI conversations ### Chat conversations (Web, SMS, Email, Facebook, Instagram, Google Business Messages, WhatsApp) All chat-based AI conversations flow into your **Inbox** just like human-handled conversations. Open **Inbox** from the sidebar and you'll see the threads listed alongside everything else. AI-handled threads look like any other conversation; the message history shows the AI's replies inline. Filter the inbox by channel or by tag to narrow down to AI-only threads. See Filtering Conversations. ### Voice AI transcripts Voice AI calls are stored under the **Voice / Calls** area of your account. Each call has: * A recording (if recording was enabled at call time) * A written transcript of the conversation * Any lead data the AI captured (name, phone, email, appointment interest) * The contact profile the AI linked the call to To find a specific Voice AI conversation, open the contact profile and look at the **Activities** tab — the call and its transcript are logged there. To browse all Voice AI calls in one list, use the Voice section directly. If you're not seeing Voice AI transcripts: * **Check that Voice is enabled** as a deployment channel under Configure AI → Deploy AI to Channels. * **Confirm the number the AI answered has Voice enabled** on your account. * **Contact** [**support@gleantap.com**](mailto:support@gleantap.com) if calls are happening but transcripts aren't showing up. *** ## Handoff to a human The **Transfer to Human** toggle under Configure AI Chatbot lets the AI hand off a conversation to your team — either because the visitor asked for a person or because the AI detected it can't help. When configured well, the visitor sees a message like "*Let me bring in someone from my team to help you with this*", and the conversation lands in your Inbox for a teammate to pick up. ### What happens under the hood When the AI triggers a handoff: 1. **The visitor receives your transfer message** (configured under Configure AI → Transfer to Human). 2. **The conversation is escalated in the Inbox** — it moves to an active/open state waiting for a human reply. 3. **Notifications fire** to the teammates you configured under handoff settings. This can be specific users or entire roles (Admin, Workroom Admin, Sales, Marketing, Analyst). 4. **Outside working hours** — if the handoff request comes in outside the "Human Working Hours" you set, the visitor gets your configured quiet-hour message instead (e.g., "*Our team is offline until 9am. We'll get back to you first thing.*"), and no notification is sent to the team. ### Working hours behavior Working hours are configured under **Configure AI → Transfer to Human**: * **Within working hours** — full handoff runs; team gets notified; visitor gets the transfer message. * **Outside working hours** — visitor gets the quiet-hour message; no team notification (they're off the clock). If you want notifications 24/7, extend working hours to cover the full day. *** ## Troubleshooting: "The AI said 'someone will reach out' but no one did" This is the most common AI issue. There are three failure modes; walk through in order. ### 1. No handoff recipients configured Open **Configure AI → Transfer to Human** and confirm you've selected at least one role or specific user to receive handoff notifications. If the recipient list is empty, the handoff runs silently — the visitor sees the transfer message, but no one on your team gets pinged. **Fix:** add at least one role (Admin, or a Sales/Marketing role you've populated with users) or select specific teammates. ### 2. The handoff happened outside working hours Check the timestamp of the handoff against your configured **Human Working Hours**. If the handoff was outside those hours, the system intentionally suppresses the team notification — the visitor gets the quiet-hour message you configured, and you'll pick it up when you come online. **Fix:** either extend working hours, or make sure the quiet-hour message tells the visitor when to expect a reply so they know what's happening. ### 3. The teammate's Inbox Notifications are off Even if the AI correctly assigns the conversation to a teammate, that teammate must have Inbox notifications enabled — web push, email, or SMS — to hear about it. If notifications are disabled, the conversation lands silently in their queue and stays there until they check the Inbox manually. **Fix:** open **Configure → Notifications** and confirm each teammate has at least one notification channel enabled. See Notifications and SMS Notifications. ### Quick verification To verify handoff end-to-end: 1. Enable Transfer to Human with your own user as the recipient. 2. Open the chat widget in an incognito window and trigger a handoff (e.g., ask to speak to a person). 3. Confirm you receive the notification through your configured channel(s) within a minute. If step 3 fails, walk through the three fixes above. *** ## Turning the AI off You have three levels of control: **Turn off entirely** — under **Configure AI Chatbot**, toggle **Enable AI Agent** off. The agent stops handling any conversations on any channel. Inbound messages arrive in the Inbox and wait for a human. **Turn off for one channel** — under **Deploy AI to Channels**, uncheck the channel(s) you want the AI to skip. For example, keep the web widget on but disable SMS if you want the AI answering website visitors but not text messages. **Turn off Voice AI specifically** — uncheck **Voice** in the deploy channels. The AI will stop answering inbound calls; those calls will go to whatever voice routing you have configured (voicemail, forward, IVR). *** ## Common scenarios ### "The AI booked an appointment but it didn't show up in my calendar" Check the Tools configuration — Book Appointments must be enabled and appointment types must be set up. Booked appointments appear in **Appointments** and sync to Google/Outlook if you've connected those calendars. See Sync with Google & Outlook. ### "The AI is answering questions I want a human to handle" Update your **Custom Intents** to route those questions differently. You can configure an intent so the AI recognizes the question but responds with the handoff message rather than answering itself. See Custom Intents. ### "The AI is giving wrong answers about pricing or hours" Retrain it. Upload the correct info in **Train AI** — via CSV, website URL, or free-text notes. The AI uses that as its knowledge base for future replies. See Train AI. ### "How do I see everyone the AI talked to this week?" Build a segment in **Audience → Segments** filtering on the source or tag your AI Agent sets, plus `Added On is in last 7 days`. See Filter Options. ### "The AI keeps replying to reactions on SMS (thumbs up, heart)" Known limitation. Reach out to if this is a recurring problem for you — we can advise on a workaround for your setup. *** ## Related pages * Configure AI Chatbot — initial setup * Train AI — knowledge base * Custom Intents — Q\&A pairs * Tools — appointment booking * Test Your AI Chatbot — preview before enabling * Embed Your AI Chatbot Widget — website installation * Notifications — how team notifications work --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://help.gleantap.com/ai-chatbot/managing-ai-conversations.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.