AI voice agent that qualifies leads and books appointments

The problem

The core problem was that handling calls, qualifying leads, and booking appointments manually could not cover calls at all hours or keep response times consistent.

• Slow lead response: Leads calling or messaging outside business hours had no way to be engaged immediately, risking lost opportunities.
• Inconsistent qualification: Manually screening callers for fit and intent varied by staff member and time of day, with no standardized scoring process.
• Manual appointment scheduling: Booking meetings required a human to check availability and coordinate directly with the caller in real time.
• Repetitive support inquiries: Common customer questions consumed staff time that could otherwise go toward higher-value work.
• Disconnected call records: Conversation outcomes and lead status lived in call logs disconnected from the CRM, requiring manual updates.
• No unified visibility: Teams lacked a single place to see call activity, lead status, and follow-up needs across both voice and WhatsApp channels.

What we built

The agent was built around real-time voice and messaging infrastructure, with AI driving the actual conversation, qualification judgment, and routing decisions throughout each interaction.

• Call and message intake (fixed infrastructure): Twilio provisions a dedicated US phone number for inbound and outbound calls, paired with WhatsApp API integration for messaging-based interactions.
• Real-time conversational AI (AI-driven judgment): VAPI and OpenAI power natural, context-aware conversation for every call, allowing the agent to understand caller intent and respond dynamically rather than following a rigid script.
• Lead qualification (AI-driven judgment): During the call, the agent asks screening questions and scores the caller’s responses to determine lead quality, a genuine judgment call rather than a fixed checklist.
• Routing decision (AI-driven judgment): Based on the qualification score, the agent routes hot leads differently from lower-priority ones, directing the conversation or follow-up accordingly.
• Appointment booking (AI-driven action): When appropriate, the agent books a meeting directly into the calendar during the live call, without requiring a human to confirm availability afterward.
• Customer support handling (AI-driven judgment): The agent resolves common inquiries conversationally and identifies when a request requires escalation to a human.
• Receptionist duties (AI-driven judgment): For general callers, the agent greets, collects relevant information, and determines the appropriate person or department to route the caller to.
• Workflow orchestration (fixed logic): N8N coordinates the backend logic connecting call events, qualification outcomes, and appointment actions to downstream systems.
• CRM sync and dashboard logging (fixed logic): Every call’s conversation data, lead status, and outcome are automatically synced to a live dashboard for centralized team visibility.

How it works

The system runs on the tools listed below. Platform names appear here and in the stack, so you can see exactly what was used.

• Voice infrastructure: Twilio for phone number provisioning and call handling
• Conversational AI: VAPI for real-time voice agent orchestration
• AI/LLM: OpenAI for natural language conversation, qualification, and decision-making
• Messaging: WhatsApp API for message-based interactions
• Orchestration: n8n workflow automation for backend logic and integrations
• Dashboard and CRM: Retool for centralized call activity, lead status, and follow-up management