The Implementation Guide: Build a WhatsApp Revenue Agent in 15 Mins

So you've bought into the strategy. Now let's build the machine. Follow this guide to go from "Blank Screen" to "Live Agent" on Peach AI without writing a single line of code.

Step 1: Initialize Your Agent

First, we need to choose the right architecture. Peach offers pre-built templates, but for maximum revenue impact, we want full control.

1. Log in to your Peach Dashboard.

2. Navigate to the AI Agents tab and click New AI Agent.

3. You will see three options. Select "Build Your Own Agent".

   
   

   

Step 2: Define the "Brain" (The Prompt)

This is where you give your agent a personality. Do not use generic prompts like "You are a helpful assistant." Be specific.

1. Go to the Instructions tab.

2. In the Goal section, define the high-level objective.

   • Example: "Your objective is to help the user achieve their stated goal. Persuade and motivate the user by highlighting benefits."

3. In the Conversation section, define the persona.

Copy/Paste this Template:"You are [Name], a Product Specialist at [Company]. Your goal is to resolve questions using the Knowledge Base and identify customers ready to buy (Warm Leads). Tone: Simple, confident, and enthusiastic. Act like a smart friend. Rule: Do not send long paragraphs. Keep responses short and crisp."

Step 3: Configure Reporting & Data Capture

We don't just want a chat; we want structured reporting data. We need to tell the agent how we need the output report from the customer conversations.

1. Go to the Reporting tab.

2. Define the Fields you want to extract from the actual conversations.

   • Example: ‘Intent Tag: SQL (if they want to buy now) or MQL (if they are just exploring)’

   • Purchase Timeline: Capture values like "This week" or "Not sure".

   • Objections: Capture hesitation points (e.g., "Price too high").

Step 4: Upload the Knowledge Base (RAG)

Your agent needs to know your business. Peach gives you two ways to train it: the Universal Library (for general info) or the Agent Notebook (for specific info).

Option A: The Universal Knowledge Base (Efficient) If you have a general "Company Policy" or "Product Catalog" that all your agents (Sales, Support, Triage) need to know, upload it once here.

1. Navigate to Knowledge Bases in the main left-hand sidebar (outside the agent builder).

2. Click New Knowledge Base and upload your documents (PDF, DOC, TXT).

The Benefit: Any future agent you build can automatically "subscribe" to this library without you needing to re-upload files.

Option B: The Agent-Specific Knowledge Base (Focused) If this specific agent needs private data (e.g., a "Sales Playbook" that the Support Agent shouldn't see), add it directly inside the builder.

1. Inside the Agent Builder, go to the Knowledge Base tab.

2. Select an existing Universal Base OR upload a new file specific to this agent.

The Magic: Peach automatically "chunks" this info so the AI can search it instantly. If a user asks, "Is this machine washable?", the agent looks up the manual and answers, "Yes, but only on a cold cycle".

Step 5: Give It Hands (Connect Tools)

This is the difference between a dumb chatbot and a smart agent. Let's give it permission to do things.

1. Go to the Tools tab. You need your agent to connect to your apps. Peach makes this "No-Code" via MCPs or CRM connectors.

2. Click Connect next to the integration you need:

   • Shopify: To check real-time stock levels.

   • HubSpot / Freshsales: To push "Hot Leads" directly into your pipeline.

   • WhatsApp: To enable native messaging features.

Step 6: Set the Safety Net (Escalation)

You never want a high-value customer to feel stuck. Set up the "Human Handoff" rules.

1. Go to the Escalation tab.

2. Define Conditions: Tell the AI when to give up.

   • Example: "When a customer requests a refund, transfer the conversation to a human."

3. Select Assignee: Choose the specific human agent (e.g., Dharmin Patel) who should receive the escalation.

4. Escalation Mode: Choose "Handoff To Human" (AI stops talking) or "Hybrid" (AI stays to help).

Step 7: Test Before Launch

Never launch blindly. Use the simulator to try to "break" your agent.

1. Use the Test Widget on the right side of the screen.

2. Run scenarios (+New Chat):

   • Scenario A: Ask about a specific product feature ("What about suction?").

   • Scenario B: Test the logic ("I have a small kitchen").

3. Watch the Analysis logs to see how the agent analyses your chat.

Step 8: Go Live

Once satisfied, it's time to deploy.

1. Click Publish.

2. Choose your Launch Strategy:

   • Option 1: Launching a WhatsApp broadcast and linking your AI agent (reply automation) 

Option 2: Linking AI agent to respond to a CTWA campaign

You can create an automation by selecting a default Agent to respond to any incoming customer messages on WhatsApp