An AI chatbot human handoff is not complete when the assistant says, “I’ll get someone.” It is complete when the right person receives the request, understands what remains unresolved and knows what to do next.
Before launching a website assistant or WhatsApp chatbot, use this checklist to define that process. Treat it as an operating agreement between the people maintaining your content and the staff handling escalations—not just a chatbot setting.
1. Decide where automated help should stop
List the questions your assistant can answer from approved information. Then separate requests that require judgment, additional permission or information it cannot access.
For example, explaining a published cancellation policy is different from approving an exception. Describing how to update account details is different from changing them.
Document these boundaries explicitly. QuerySafe’s permissions and human approval approach distinguishes answering questions from controlled actions; action permissions and approvals apply where Agent capabilities are enabled. Do not assume a customer-facing conversation automatically authorizes a system change.
2. Write an escalation decision list
For each trigger, record the expected response, destination and fallback. Start with these cases:
- The customer asks for a person: acknowledge the request and follow the agreed transfer process rather than forcing more automated questions.
- The answer is missing: explain the information gap and pass along the unanswered question.
- The customer disputes a policy: preserve the issue for someone authorized to review exceptions.
- The request concerns sensitive account information: use your approved verification process before disclosing details or making changes.
- The conversation repeats without progress: stop rephrasing the same answer and offer the defined next step.
Make each rule observable. “Escalate when appropriate” is harder to test than “After an unsuccessful clarification, offer staff review.”
3. Define the minimum handoff brief
A useful brief should let staff continue without asking the customer to restart. Specify:
- The customer’s goal.
- Relevant conversation history or a concise summary.
- What has already been answered or attempted.
- The unresolved question and reason for escalation.
- The next action and responsible team.
Include only necessary personal information. Avoid collecting passwords, payment credentials or unrelated documents to make a handoff feel complete.
QuerySafe for WhatsApp supports routing conversations to staff with history and context preserved. Confirm the exact routing and brief requirements for your workflow during scoping.
4. Prepare honest transition messages
Write messages for different staffing conditions. These are templates to adapt, not promises to enable without an agreed process:
Staff available: “This needs a team member’s help. I’ll pass along your question and what we’ve discussed.”
Outside staffed hours: “The team is currently offline. I can collect your question for review during staffed hours.”
No immediate resolution: “I don’t have enough approved information to answer that. A team member will need to review it.”
Only promise response times your team can meet. Make any alternative contact method visible and verify that it is monitored.
5. Assign ownership and a fallback
Name a primary owner and backup for each request category. Decide who checks unattended conversations, what happens when the assigned person is unavailable and when staff should take control of the conversation.
Also define whether automated replies should pause during human handling and how the conversation returns to automated help. Confirm that the proposed channel setup supports the behavior you need.
6. Test before launch, then review failures
Create a test sheet with columns for the scenario, expected outcome, actual outcome, owner and fix. Include direct requests for a human, ambiguous wording, repeated questions, language changes, missing information and unavailable staff.
Check that the destination receives enough context—not merely that the chatbot displays a transfer message. Resolve failures that could leave customers without a usable next step before launch.
After launch, review missed escalations, unnecessary transfers and unanswered questions. Update approved content where appropriate, but retain human review for genuine exceptions.
Bring your test sheet to a QuerySafe demo. Confirm website and WhatsApp handoff requirements separately rather than assuming identical capabilities across channels.
