Seamless Human Handoff for WhatsApp
When AI isn't enough, empower your team to take over conversations instantly. Here's how our handoff system works.
Why Handoff Matters
Even the most capable AI will encounter questions it should not answer alone — complex complaints, sensitive account issues, or situations where empathy matters more than speed. A clean handoff experience preserves customer trust and gives your agents the context they need to help immediately.
The WhatsApp Bot Platform uses a three-state conversation model: bot, human, and closed. Transitioning between states is instant and auditable, with a full message history visible to every agent who picks up the conversation.
Triggering a Handoff
Handoff can be triggered in three ways. First, the customer can type a keyword you configure (e.g. 'speak to agent'). Second, a Flow step with type 'Human Handoff' transfers control when a specific path is reached — for example, after a failed self-service attempt. Third, the LLM itself can request a handoff by calling the built-in handoff tool when it determines it cannot resolve the query.
When handoff triggers, the platform publishes a real-time event to the agent dashboard. All online agents for that tenant see a notification with the conversation preview and can claim it with a single click.
The Agent Experience
Agents work inside the Conversations view. The left panel lists open conversations grouped by status. Clicking a conversation opens the full message history — including any variables the flow collected and any tool call results. Agents type replies directly in the chat panel and send them via the platform's WhatsApp gateway.
When the issue is resolved, the agent marks the conversation as Closed. The customer receives a configurable closing message, and the conversation is archived for reporting.
Monitoring and SLA Tracking
The Analytics page (coming soon) will surface median handoff response time, resolution time by agent, and escalation rate — giving team leads the data they need to right-size staffing and improve bot coverage over time.
For now, tenant admins can filter Conversations by status and date to get a manual view of queue depth and agent activity. The conversation timestamps make SLA breaches visually obvious.