Skip to content
SquadOS SquadOS Docs

AI Agent for WhatsApp Customer Support

SquadOS lets you build an AI agent that automatically responds on WhatsApp — 24/7 — using your company’s context.

What you’ll achieve

Section titled “What you’ll achieve”

By the end of this guide, you’ll have an agent that:

  • Answers your customers’ frequently asked questions on WhatsApp
  • Queries your knowledge base (FAQs, manuals, policies) to respond accurately
  • Transfers conversations to a human agent when needed
  • Logs all conversations for later analysis

Why use SquadOS for WhatsApp support?

Section titled “Why use SquadOS for WhatsApp support?”

Unlike traditional chatbots with fixed flows, SquadOS uses language models (GPT-4, Claude, Gemini) that understand user intent and respond naturally. You define agent behavior through a system prompt — not menus or decision trees.

Step by step

Section titled “Step by step”

  1. Create the agent

    In the admin panel, go to Agents → + New Agent. Give it a clear name like “WhatsApp Support” and click Create Agent.

  2. Configure the system prompt

    In the editor, open the Prompt tab and define:

    • The agent’s role (“You are the support assistant for [Company]…”)
    • The conversation tone (formal, casual, friendly)
    • What the agent can and cannot answer
    • When it should transfer to a human

    Use AgentMaker (the “Edit with AgentMaker” button) if you need help writing the prompt.

  3. Connect your Knowledge Base

    If you have documents (FAQs, manuals, usage policies), create a Knowledge Base and link it to the agent in the Tools tab. The agent will automatically query those documents when answering.

  4. Configure WhatsApp

    In the editor’s Triggers tab, choose:

  5. Test before publishing

    Click Test Agent and simulate real customer questions. Adjust the prompt until you’re satisfied with the responses.

  6. Activate and monitor

    Once the agent is published, all conversations appear in Conversations. Use Analytics to track volume, satisfaction, and token usage.

Section titled “Recommended settings for customer support”
SettingRecommended valueReason
ModelGPT-4o mini or Claude HaikuCost/quality balance for support
Human transferEnabledAlways provide a human escalation path
GuardrailEnabledPrevents out-of-scope responses
Context history10-20 messagesMaintains coherence without excessive cost

Next steps

Section titled “Next steps”