Kaizen 233 Generating AI-Powered Follow-up Emails Using CRM Functions and Widgets

Hey everyone!

Welcome back to another interesting post in the Kaizen series!

Sales teams regularly capture interaction notes in CRM after speaking with prospects. However, drafting a follow-up email that reflects the conversation context can be repetitive and time-consuming.

What if CRM could automatically generate a contextual follow-up email draft based on recent interactions with the lead?

In this Kaizen post, we will create a practical tool that generates context-aware follow-up emails directly from a Lead record. You will see how it pulls recent notes, lead's details, and crafts a ready-to-send draft.

The solution uses three Zoho strengths:

Widgets for intuitive UI
Functions for secure backend processing
External AI APIs (like Groq or OpenAI)

The output displays in the widget. Review, tweak if needed, copy, and paste into your email composer.

The business challenge

Sales reps log calls like "Prospect interested in 20 user licenses but concerned about pricing. Explore volume discounts."

Now, crafting a response involves recalling details, matching tone, and ensuring professionalism. Miss a nuance, and trust erodes.

Our AI generator automates this. It scans the lead's details (name, company, status) and the three most recent notes, then prompts an AI model for a draft.

Example output

Subject: Follow-up on CRM Licensing Discussion

Hi John,

Thanks for the insightful chat today about your team's CRM needs. You're considering licenses for around 20 users, and I appreciate pricing being a top priority.

I'll review our volume options and discount eligibility right away. Expect specifics by EOD tomorrow.

Best regards,

[Your Name]

This keeps the rep in control while slashing draft time from 10 minutes to 10 seconds. Scalable for high-volume teams.

High-level architecture

Think of a layered cake!

UI Layer (Widget): Button launches it; displays, copies the draft, and sends the email.
Logic Layer (Function): Fetches CRM data server-side, builds prompt, calls AI securely.
AI Layer: External service generates natural language.

Flow

Rep clicks a button → Widget gets Lead ID → Calls function → Function queries CRM + AI → Draft back to widget

Secure (no client-side keys), reliable (server-side), and reusable across modules.

Developers often wonder "Why not fetch AI responses straight from widget JavaScript?"

Widgets run in a browser iframe, so embedding API keys exposes them to inspection or misuse. CORS policies can also block external requests, and complex logic like prompt building suits server-side better.

So, the best practice is to use widgets for UI and functions for secure server-side work. This hides credentials and reuses logic.

Follow these steps to build this solution.

Step 1: Add the custom Button

Head to Setup → Modules and Fields → Leads → Buttons → Create New Button.
Provide the following details:

Name: Generate AI Email
Placement: Record Detail Page (top or bottom)
Action: Open Widget (select your widget once built)

Save and add to layout.
Test: Button appears on Leads; clicking loads widget in context.

Pro tip: Position near the "Send Email" button for seamless workflow.

Step 2: Build the CRM Function

Go to Setup → Developer Hub → Functions → + Create Function.
Enter the display name, function name, description, choose Standalone as the category.
Click Create. The Functions IDE opens.

Enter the following function code. Note that this example uses Groq AI API.

string standalone.generate_ai_email(String leadId)
{

lead = zoho.crm.getRecordById("Leads",leadId);

name = ifnull(lead.get("Full_Name"),"");

company = ifnull(lead.get("Company"),"");

status = ifnull(lead.get("Lead_Status"),"");

// Fetch recent CRM Notes

notesResp = zoho.crm.getRelatedRecords("Notes","Leads",leadId);

recentNotes = "";

count = 0;

for each note in notesResp

{

noteContent = ifnull(note.get("Note_Content"),"");

if(noteContent != "")

{

recentNotes = recentNotes + "[" + (count + 1) + "] " + noteContent + ". ";

}

count = count + 1;

if(count == 3)

{

break;

}

// Build AI prompt

prompt = "You are a sales assistant helping draft follow-up emails. ";

prompt = prompt + "Generate a professional follow-up email based on the customer's previous interactions. ";

prompt = prompt + "Lead Name: " + name + ". ";

prompt = prompt + "Company: " + company + ". ";

prompt = prompt + "Lead Status: " + status + ". ";

prompt = prompt + "Recent CRM Interaction Notes: " + recentNotes + ". ";

prompt = prompt + "Write a professional follow-up email referencing the discussion.";

// Sanitize prompt

prompt = prompt.replaceAll("\"","\\\"");

prompt = prompt.replaceAll("\n"," ");

prompt = prompt.replaceAll("\r"," ");

// Build AI request

message = Map();

message.put("role","user");

message.put("content",prompt);

messages = List();

messages.add(message);

requestBody = Map();

requestBody.put("model","llama-3.1-8b-instant");

requestBody.put("messages",messages);

requestBody.put("temperature",0.4);

requestJSON = requestBody.toString();

// Call Groq API

response = invokeurl

[

url :"https://api.groq.com/openai/v1/chat/completions"

type :POST

parameters:requestJSON

headers:{"Authorization":"Bearer YOUR_API_KEY","Content-Type":"application/json"}

];

// Extract AI email

emailContent = "";

if(response.containsKey("choices"))

{

emailContent = response.get("choices").get(0).get("message").get("content");

}

// Return generated email to widget

return emailContent;

}

Note for production

Connections: In the example, the API key is included directly in the function for simplicity. However, production deployments should use Connections instead of hard-coding credentials as they allow developers to store API credentials securely within Zoho CRM and reuse them across multiple integrations.
Error Handling: Wrap invokeurl in try-catch.
Limits: Cap notes at 3 to avoid token overflow and tune temperature (0.4 = consistent).
Logging: Add info prompt before API call for debugging.
Test: Execute with sample Lead ID. Expect clean email string back.
Extend: Add phone/email from Lead, or filter notes by date (e.g., last 7 days via criteria).

Step 3: Develop the Widget

The widget provides the user interface for generating and copying the AI email.

When the widget loads, it retrieves the current Lead ID using the CRM widget JS SDK. When the user clicks Generate Email, the widget calls the CRM function and displays the generated email draft.

The widget also provides Send Email and Copy Email buttons that send the email to the lead record's email ID and copies the email content to the clipboard so the rep can past it into the email composer, respectively.

Sample index.html

<!DOCTYPE html>

<html>

<head>

<meta charset="UTF-8">

<script src="https://live.zwidgets.com/js-sdk/1.3-beta/ZohoEmbededAppSDK.min.js"></script>

<style> 

</style>

</head>

<body>

<div class="widget-card">

<div class="widget-header">

<svg width="16" height="16" viewBox="0 0 24 24" fill="none" xmlns="http://www.w3.org/2000/svg">

<path d="M20 4H4C2.9 4 2 4.9 2 6v12c0 1.1.9 2 2 2h16c1.1 0 2-.9 2-2V6c0-1.1-.9-2-2-2zm0 4l-8 5-8-5V6l8 5 8-5v2z" fill="white"/>

</svg>

<h2>AI Follow-up Email Generator</h2>

</div>

<div class="widget-body">

<div class="email-box-wrapper">

<textarea id="emailBox" placeholder="Click 'Generate Email' to create a contextual follow-up email..."></textarea>

</div>

<div class="action-bar">

<button class="btn btn-primary" id="generateBtn" onclick="generateEmail()">

<svg width="13" height="13" viewBox="0 0 24 24" fill="none"><path d="M12 2l2.4 7.4H22l-6.2 4.5 2.4 7.4L12 17l-6.2 4.3 2.4-7.4L2 9.4h7.6z" fill="white"/></svg>

Generate Email

</button>

<button class="btn btn-success" id="sendBtn" onclick="sendEmail()">

<svg width="13" height="13" viewBox="0 0 24 24" fill="none"><path d="M2 21l21-9L2 3v7l15 2-15 2v7z" fill="white"/></svg>

Send Email

</button>

<button class="btn btn-secondary" onclick="copyEmail()">

Copy

</button>

</div>

<div class="status-bar" id="statusBar"></div>

</div>

<script src="index.js"></script>

</body>

</html>

Sample index.js

let leadId;

let leadEmail;

let generatedEmail = "";

// Widget initialization

ZOHO.embeddedApp.on("PageLoad", function(data){

if(data && data.EntityId){

leadId = data.EntityId[0];

// Fetch Lead email

ZOHO.CRM.API.getRecord({

Entity: "Leads",

RecordID: leadId

})

.then(function(response){

if(response && response.data && response.data.length > 0){

leadEmail = response.data[0].Email;

}

});

}

});

ZOHO.embeddedApp.init();

// --- Helper functions used by email generation/sending ---

// function formatEmailText(raw) { /* formatting logic */ }

// function textToHtml(text) { /* convert plain text to HTML */ }

// function setStatus(message, type) { /* UI status logic */ }

// Generate AI email using a CRM function

function generateEmail(){

var req_data = {

arguments: JSON.stringify({

leadId: leadId

})

};

ZOHO.CRM.FUNCTIONS.execute('generate_ai_email', req_data)

.then(function(response){

if(response && response.details){

generatedEmail = formatEmailText(response.details.output);

document.getElementById('emailBox').value = generatedEmail;

}

})

.catch(function(error){

console.log('Function error:', error);

});

}

// Send email using ZRC

async function sendEmail() {

const emailContent = document.getElementById('emailBox').value.trim();

if (!emailContent) return;

try {

// Get allowed "From" addresses

const fromRes = await zrc.get('/crm/v8/settings/emails/actions/from_addresses');

const fromAddress = fromRes.data.from_addresses[0];

const htmlContent = textToHtml(emailContent);

// Send mail

const response = await zrc.post(`/crm/v8/Leads/${leadId}/actions/send_mail`, {

data: [

{

from: {

user_name: fromAddress.display_name || fromAddress.user_name,

email: fromAddress.email

to: [

{

email: leadEmail

}

subject: 'Follow-up',

content: htmlContent,

mail_format: 'html'

}

]

});

console.log('Mail sent:', response);

} catch (error) {

console.error('Send mail error:', error);

}

// Copy generated email

// function copyEmail(){ /* copy logic */ }

The ZIP containing the complete index.js and index.html files used in this example is attached at the end of this post.

Refer to Creating a Widget in Zoho CRM and the JS SDK for more details on CLI installation, creating, packaging, and hosting a widget.

User experience walkthrough

The salesperson clicks Generate AI Email on the record detail page and the widget opens.
The salesperson clicks the Generate AI email button. The widget retrieves CRM context and generates a contextual follow-up email draft.
The email appears inside the widget, allowing the salesperson to quickly review the generated message, and use the Send Email button to send it to the lead's email ID.

This workflow ensures that AI assists with drafting the message while still allowing the salesperson to review and control the final communication.

Here is a GIF explaining the use case in action.

Key takeaways

This example demonstrates how Zoho CRM developers can integrate generative AI into CRM workflows using platform-native tools.

Combine CRM Widgets for user interaction and Functions for secure server-side processing to safely integrate external AI services without exposing credentials or compromising security.

The same architecture can be extended to build more advanced AI features, such as:

Automated meeting summaries
Deal health insights
Proposal drafting assistants
Intelligent follow-up recommendations

We hope you liked this post. Let us know what you think in the comments or reach out to us at support@zohocrm.com.

Cheers!

===================================================================================

See Also