> For the complete documentation index, see [llms.txt](https://converso.gitbook.io/product-documentation/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://converso.gitbook.io/product-documentation/ai-agents/human-handoff.md). # Human handoff One of the most important features of Converso is the ability for conversations to move seamlessly between AI agents and human team members - and back again. This is what we call **human handoff**, and it's built into every AI agent on the platform. Unlike traditional chatbot platforms where AI and human support exist as separate, disconnected experiences, Converso operates both inside the same shared inbox. This means your team has full visibility of every conversation, whether it's being handled by the AI or by a person - and switching between the two is instant, with no context lost. ### How conversations start When an AI Agent is configured in a workspace, all new inbound conversations are automatically assigned to it. Whether the message arrives via webchat, WhatsApp, or another connected channel, the AI agent picks it up and begins responding to the customer immediately - no manual routing required. ### AI to human handoff A conversation can move from an AI agent to a human team member in two ways: the AI offers, or the customer asks. #### The AI offers a handover Your AI agent recognises when a conversation needs a person. It will offer a handover when: * the question can't be answered from your knowledge base * the request is account-specific — billing, contracts, or anything needing a person's judgement * the topic is sensitive or high-stakes, such as legal, compliance, security, or data * the customer reports a problem that has cost them something * the customer is frustrated, or repeating a question without getting anywhere In each case the agent offers to pass the conversation to your team rather than transferring on its own. If the customer accepts, the handover goes ahead. #### The customer asks for a person At any point the customer can ask to speak to someone. Any of these is enough: * "Transfer to agent" * "Speak to someone" * "Support" The agent recognises the intent and passes the conversation straight over. It won't ask them to confirm what they've already asked for, and there are no menus, buttons, or new chats to start. #### What happens after handoff When the AI agent unassigns itself from the conversation, a notification is sent to all human agents in the workspace, letting them know the chat needs an available person to take over. The chat can then be assigned manually to any human agent. That agent can review every previous message in the conversation for full context — including everything the AI discussed with the customer — and reply via live chat. The conversation also appears in that agent's My Inbox for easy tracking. The customer never has to repeat themselves, and your team picks up exactly where the AI left off.AI to human handover There are two ways a conversation can move from an AI agent to a human team member. #### When no human is available If a human agent is not assigned to the conversation within 120 seconds after handover, a customer contact form is sent to the customer via the widget, who then has the option to leave their contact details, such as phone number and email address. These details are then stored in Contacts, so your team can follow up when convenient. ### Handback to AI Handoff isn't a one-way street. Once a human team member has resolved the customer's issue, they have two options for returning the conversation to AI control. The first is to assign the conversation back to the AI agent directly. This is useful when the human has dealt with a specific query — perhaps a sensitive account issue or an unusual request — but the customer is still chatting and may have follow-up questions the AI is perfectly capable of handling. Rather than staying tied to the conversation, the human agent can hand it back to the AI and move on to the next priority. The second option is to close the chat entirely. Once a chat is closed, any new messages from that customer will be picked up and responded to by the AI agent automatically, starting a fresh conversation. **Coming soon:** The AI agent will also be able to close chats automatically when the customer indicates they're satisfied — for example, responding with "thanks" or a similar comment that signals no further help is needed. This flexibility means your team spends their time where it matters most, while the AI handles the rest. ### Monitoring and manual intervention All AI agent conversations are viewable in real time in the team inbox. Your team doesn't need to wait for a formal handover to get involved - a human agent can stop the AI from replying in any chat at any time by simply unassigning the conversation from the AI agent. The human agent can then pick up the conversation and live chat directly with the customer. This gives you complete oversight without requiring constant supervision. You're not managing two separate systems or checking a chatbot dashboard in one tab and a helpdesk in another. It's all in one view, with full control. ### Disabling the AI agent To stop the AI agent from responding to all new inbound messages across a workspace, go to Workspace **> AI Agents**, and unpublish the agent. This immediately stops the AI from picking up new conversations, while preserving all existing chat history. ### Why this matters Most businesses that deploy AI for customer conversations worry about the same thing: what happens when the AI can't help? If the answer is a dead end, a generic error message, or a frustrating loop, the customer experience suffers and the AI becomes a barrier rather than a benefit. Converso's handover model is designed to solve this. The AI handles what it can, recognises what it can't, and connects the customer to a human without friction. The human resolves the issue with full context, then the chat is closed so the AI can resume on the next interaction. It's a continuous loop that keeps customers happy and your team efficient. This approach means you get the best of both worlds: the speed and availability of AI for routine queries, combined with the empathy and judgement of your team for everything else. No black boxes, no disconnected experiences, and no customers left waiting. ### Key points * All new inbound conversations are automatically assigned to the AI agent when configured in a workspace. * The AI detects when its responses aren't resolving the customer's question and proactively offers to transfer to a human. * Customers can request a human at any point with a simple phrase like "transfer to agent" or "support". * When the AI unassigns itself, all human agents in the workspace are notified that the chat needs attention. * Human agents receive the full conversation history, so the customer never has to repeat themselves. * If no human is available, the AI captures the customer's contact details for follow-up. * Closing the chat after resolution ensures new messages from that customer are handled by the AI again. * Human agents can manually unassign the AI from any conversation at any time to take over directly. * The AI agent can be disabled entirely via workspace > AI Agents.