AI FAQ + Human Handoff

Run a Telegram support bot that answers from a knowledge base, routes order and return requests, and hands off complex conversations to a human.

The user wants the same support flow as the AI support bot article: a knowledge-base AI assistant, explicit routes for order status and returns, and a safe handoff path for human support.

• Which order, delivery, return, and warranty questions should the AI assistant answer directly?
• Which topics should always go to a human?
• Should order and return requests stay as placeholders for staff review or be created in an external system later?
• Before going live, review the AI FAQ Assistant prompt and add real business information, support topics, policies, pricing rules, account rules, and escalation rules.
• Replace the seeded demo knowledge base with the business's real FAQ and support content.
• If support requests must be created in a real system, replace the placeholder request action with an approved helpdesk, CRM, ticketing, or form integration.
• To alert your team about each new request, tag staff contacts on the Contacts page with the team tag you picked during setup (default "support-staff"). The Notify support staff action then messages everyone carrying that tag with the request summary and follow-up email.

• Start with a welcome message that explains the bot can answer order, delivery, return, and warranty questions
• Use an AI_TOOL_ROUTER FAQ assistant in INTENTS_AND_CHAT mode
• Let the AI answer common support questions from the attached knowledge base
• Route order-status intent to an order-number collection flow
• Route return-or-refund intent to a return guidance flow
• Route staff-handoff intent to START_CONVERSATION
• Use AI_ACTION to summarize captured issue details for support staff
• Use placeholder request ID/status until a helpdesk, CRM, or order system integration is connected

Good to know

• Use AI_TOOL_ROUTER with INTENTS_AND_CHAT for the main FAQ conversation so the bot can answer directly and route known intents.
• The template includes a seeded demo knowledge base and attaches it to the AI router; replace it with the business's real FAQ, policy, product, and support documents before going live.
• Use START_CONVERSATION for human handoff; it is a built-in action and does not require an external helpdesk system.
• Use placeholder SET_VARIABLE actions for support request ID/status unless the user explicitly provides or approves a helpdesk, CRM, ticketing, or form API.
• Use HTTP_REQUEST for Zendesk, Intercom, HubSpot, Freshdesk, Jira Service Management, Google Sheets, or another system only after the user provides or approves endpoint/auth/docs/config.
• Do not ask for API credentials in the briefing. External system setup belongs in module/API configuration.

Keep in mind

• Do not make users type command words for common support actions; rely on AI chat and intent routing.
• Do not create a long menu-first support tree when AI_TOOL_ROUTER can answer and route from one conversational entry point.
• Do not pretend order status, delivery status, refund status, or support request status is live unless an approved system API or module is connected.
• Do not use HTTP_REQUEST for helpdesk, CRM, order, or refund updates unless the user provides or approves endpoint/auth/docs/config.
• Do not let AI answer sensitive, regulated, payment-specific, or high-risk requests as facts; route those to a human or approved system integration.