Build WhatsApp Agents with Handoff, Outreach & Media Analysis (No-Code)

Key Concepts

AI Agents, WhatsApp Business API, Human Handoff, Outbound Messaging, Inbound Messaging, Knowledge Integration, Tool Usage, Media Interpretation, Relevance AI, Make.com, Pre-approved Templates, Personalized Outreach, Chat History, Customer Service, Lead Generation, Notifications, Feedback Collection.

1. Introduction and Opportunity

• The video focuses on building AI agents on WhatsApp for personalized outreach, customer response, database usage, human notification, and understanding voice, images, and documents.
• Opportunity: WhatsApp agents present a significant opportunity due to the large user base (3 billion) and business adoption (50 million). Most businesses still rely on human agents, while AI can automate many interactions.
• Problem: Existing software like Agentive, Chatbase, and ManyChat lack vital features such as human handoff, outbound messaging, chat management, CRM integration, and media processing.
• The video provides a setup that addresses these business needs, including a demo and step-by-step breakdown, with an option for setup without the official WhatsApp Business API.

2. Market Size and Engagement Metrics

• Market Size: 50 million businesses use WhatsApp, with 3 billion users globally.
• Engagement Metrics:
  • 98% open rates on WhatsApp vs. 21% on email.
  • WhatsApp messages read within 15 minutes on average.
  • 60% click-through rates on WhatsApp vs. 3% on email.
  • 68% of users prefer WhatsApp for contacting businesses.
• Regional Differences: WhatsApp adoption varies by region, with Europe, South America, and Asia being strong markets. The US has lower adoption.

3. Benefits of AI WhatsApp Agents

• 24/7 availability.
• Faster response times.
• Improved customer support.
• Handling more inquiries.
• Cost reduction.
• Staff redeployment to strategic tasks.

4. Why Agencies/Freelancers are Needed

• Setting up WhatsApp agents is technically challenging.
• Requires customized solutions and integration with specific CRMs.
• Knowledge base import and AI expertise are necessary.

5. Common Use Cases

• Customer service.
• Inbound sales and appointment setting (common with Facebook/Instagram ads).
• Marketing and lead generation (high open rates, but requires strategic approach to avoid spam).
• Notifications (order status, payment confirmations).
• Feedback and survey collection.

6. Common Niches

• Retail and e-commerce (online stores, local shops).
• Local services (salons, gyms, repair shops).
• Hospitality and travel (hotels, restaurants).
• Education and training (schools, online courses).
• Real estate and property management.
• Professional services (law firms, accountants).

7. Key Requirements for Businesses

• Human Handoff: Ability for human agents to jump in when AI lacks answers, for booking, or for complex issues.
• Outbound Messaging and Notifications: Companies need to initiate conversations for notifications, sales, and marketing. Many third-party softwares lack this.
• Chat Interface and History Access: Companies need to review chat history and intervene when necessary.
• Knowledge Integration: AI agents need access to a knowledge base to answer company-specific questions.
• Tool Usage: Integration with CRMs and other tools for updating data and accessing information.
• Media Interpretation: Understanding voice messages, images, and documents.

8. Software Options and Recommendations

• Types of WhatsApp Agents:
  • Inbound + Outbound.
  • Outbound Only.
  • Inbound Only.
• Software Options:
  • Third-Party Providers (Meta Official Partners): Agentive, Chatbase, ManyChat, Drxy, Botpress.
    • Upsides: Easy deployment, official WhatsApp Business API, knowledge integration.
    • Downsides: Usage costs, lacking key features (human handoff, outbound messaging, media interpretation).
  • Official WhatsApp Business API (Direct Integration): Naden, Relevance AI, Voiceflow, Make, Vector Shift, Flowise.
    • Upsides: Flexible AI agent builders, media interpretation, outbound messaging (with pre-approved templates), lower cost.
    • Downsides: Pre-approved messages, setup required, separate number needed, limited chat UI/handoff features.
  • Without Official WhatsApp API (Hacky Way): Relevance AI.
    • Upsides: No API setup, keep existing number, access to WhatsApp Web, 100% personalized outreach, cheapest.
    • Downsides: No media interpretation, slower response, potential errors, risk of number blocking.
• Recommendations:
  • Outbound Only: Without official WhatsApp API (Relevance AI) to keep current number and personalize outreach.
  • Inbound Only: Agentive (if no human handoff needed), or Naden/Make/Voiceflow for human handoff.
  • Complete Solution (Inbound + Outbound): Relevance AI and Make.com for media interpretation, chat history, and human handoff.

9. Detailed Setup: Outbound Only (Relevance AI, No Official API)

• Advantages: Keep current number, use WhatsApp Web, 100% personalized outreach.
• Use Case: Travel agency re-engaging past clients with personalized promotions.
• Workflow:
  1. Data Source: Google Sheets with client name, past holiday, number.
  2. Knowledge Search (RAG): Find similar destinations with promotions.
  3. Personalized Message: Send via WhatsApp.
  4. AI Agent Takeover: Respond to customer inquiries.
  5. Tools: Send WhatsApp message, deals database, general knowledge base, human handoff (delegate to manager), save answer to knowledge base.
• Tool Breakdown (Relevance AI):
  • Personalized WhatsApp Generator Tool:
    • Inputs: Price, destination, name, phone number, holiday type.
    • Knowledge Search: Find similar holiday destinations.
    • LLM Step: Select the best destination from search results.
    • Send WhatsApp Messages (3 messages with delays).
  • Retrieve Chat History: Get last five messages for context.
  • Used Travel Deal Search: Query database for deals.
  • Send WhatsApp Message: Send message to customer.
  • Knowledge Retrieval: Answer general FAQs.
  • Save Answer to Knowledge: Update knowledge base with new information.
• Human Handoff: Triggered when no answer is found or intent to book is detected. Notifications via Slack and email.
• Bulk Outreach: Upload CSV with client data and run the tool on the entire list.

10. Detailed Setup: Complete Solution (Relevance AI + Make.com, Official API)

• Use Case: Travel agency with inbound and outbound needs, including media interpretation.
• Advantages: Media interpretation, direct webhook with WhatsApp, fast response times.
• Workflow (Inbound Handling):
  1. Make.com: Receive message via webhook.
  2. Identify Message Type: Text, audio, document, image.
  3. Media Processing:
     • Audio: Voice to text (Whisper).
     • Document: Download, transcribe.
     • Image: Image analysis (LLM).
  4. Send to Relevance AI Agent:
• Workflow (Outbound Handling):
  1. Make.com: Triggered by Google Sheets data.
  2. Relevance AI (Optional): Find similar holiday destinations.
  3. WhatsApp Business Cloud Module: Send pre-approved template message with variables.
• Pre-approved Templates: Create and submit templates in Meta Business Manager.
• Agent Setup (Relevance AI):
  • Trigger: API endpoint from Make.com.
  • Core Instructions: Define agent role, goals, and escalation rules.
  • Abilities: Enable escalate to humans (email and Slack).
  • Tools: Similar to non-official setup, but with adjustments for the official API.
• Make.com Scenario Details:
  • Watch Events: Triggered by new WhatsApp messages.
  • Router: Direct messages based on type (text or media).
  • Text Message Route:
    • Send message to Relevance AI agent via API.
    • Polling Request: Check if agent is finished processing before sending new messages.
    • Sequential Processing: Process scenarios one after the other.
  • Media Route:
    • Download Media.
    • Upload to Google Drive.
    • Route based on media type (document, image, audio).
    • Process media (transcription, image analysis, data extraction).
    • Send processed data to Relevance AI agent.
• Sending WhatsApp Messages (Make.com):
  • Use "Send a Message" module (not "Send Template Message" for inbound).
  • Pass in phone number (conversation ID) and message text.
  • Option to send audio messages using 11Labs for voice synthesis.

11. Conclusion

The video provides a comprehensive guide to building AI agents on WhatsApp, covering various setups, software options, and detailed implementation steps. It highlights the significant opportunity in automating WhatsApp interactions for businesses and offers practical solutions for addressing the limitations of existing software. The combination of Relevance AI and Make.com is presented as a powerful approach for creating complete solutions with media interpretation, human handoff, and personalized outreach. The importance of understanding Meta's policies and guidelines for WhatsApp Business API is emphasized throughout the video.