Sales-to-Implementation Handoff Analyzer

What This Does

When a deal is about to close, this workflow builds a comprehensive implementation brief by pulling data from two sources — your CRM and your call recordings — then scoring how ready you actually are for kickoff.

The flow:

1. You fill out a simple form: Salesforce Opportunity ID + optional conversation history PDF

2. The workflow fetches the full opportunity from Salesforce

3. It paginates through all Gong calls, fuzzy-matches them to the account

4. It retrieves AI-generated summaries, key points, and action items from matched calls

5. It evaluates your customizable intake checklist against both SF data and call mentions

6. An LLM generates a structured pre-kickoff brief with readiness score, gaps, and recommended actions

7. The formatted brief is emailed to your implementation team

Flow

Form (Opp ID + PDF) → Extract PDF → Salesforce: Get Opportunity
  → Gong: Paginate All Calls → Match Calls to Account
  → Gong: Get Call Summaries → Build LLM Context + Intake Checklist
  → OpenAI: Generate Brief → Format HTML Email → Send to Team

Prerequisites

• n8n (self-hosted or Cloud) — needs @n8n/n8n-nodes-langchain community package for the OpenAI node
• Salesforce — Connected org with Opportunity + Account access
• Gong — API access (requires Gong Technical Admin role)
• OpenAI — API key for brief generation
• Gmail — OAuth2 for sending the brief (or swap for Slack/SMTP)

Setup (45 minutes)

Step 1: Import the Workflow

Open n8n → Workflows → Import from File → upload workflow.json.

Step 2: Configure Salesforce

1. Go to n8n Credentials → Add Credential → Salesforce OAuth2 API

2. Follow the OAuth2 flow to connect your Salesforce org

3. Assign the credential to the Salesforce - Get Opportunity node

Custom fields: The Prepare LLM Context code node maps Salesforce fields to the intake checklist. Standard fields (Name, StageName, CloseDate, Amount, Account.Name) work immediately. Custom fields (lines marked with ✏️) need to match your org's API names. Open the code node and update the sfData object (~line 10).

Step 3: Configure Gong

1. Go to n8n Credentials → Add Credential → Gong API

2. Enter your Gong API key and secret (generate in Gong → Company Settings → API)

3. In all three Gong HTTP Request nodes, replace:

- us-XXXXX in the URL with your Gong API subdomain (shown in your Gong API settings)

- YOUR_WORKSPACE_ID with your Gong workspace ID

Gong nodes to update:

• Gong - First Page
• Gong - Next Page
• Gong - Get Call Summaries

Step 4: Configure OpenAI

1. Go to n8n Credentials → Add Credential → OpenAI API

2. Enter your API key

3. Assign the credential to the Generate Brief node

4. (Optional) Change the model — default is gpt-4.1-nano for speed/cost. Use gpt-4o for higher quality.

Step 5: Configure Email

Open the Send Brief node:

• Replace implementation-team@yourcompany.com with your recipients (comma-separated)
• Connect your Gmail OAuth2 credential

Step 6: Customize the Intake Checklist

This is the most important step. Open the Prepare LLM Context code node and find the intakeChecklist array (~line 70). This defines what your implementation team needs before kickoff.

Each item has:

{
  label: 'Primary Use Case',        // What it's called in the report
  sfFields: ['primaryUseCase'],      // Which sfData key(s) to check
  gongKeywords: ['use case'],        // Keywords to scan for in call recordings
  required: true                     // Does it block kickoff?
}

To customize:

• Add/remove sections and items to match your intake process
• Update sfFields to reference your Salesforce field mappings
• Add gongKeywords that your team commonly uses in calls
• Set required: true for fields that must be filled before kickoff

Step 7: Customize the LLM Prompt

Still in the Prepare LLM Context node, scroll to the bottom to find systemPrompt and userPrompt. Update:

• The role description to match your company
• The brief sections to match your kickoff process
• Any company-specific terminology

Step 8: Test

1. Activate the workflow

2. Open the form URL (shown in the Form Trigger node)

3. Enter a real Salesforce Opportunity ID

4. Check your email for the generated brief

How the Account Matching Works

The Match Gong Calls to Account node uses fuzzy matching to find relevant calls:

1. Takes the account name from Salesforce

2. Strips common suffixes (Inc, LLC, Corp, etc.) and your custom product/brand terms

3. Tokenizes the name and scores each Gong call title by token overlap

4. Returns the top 10 matches with ≥30% token overlap (or containing the longest token)

To tune matching:

• Open the code node and add your product/brand names to the stripSuffixes function
• Adjust MATCH_THRESHOLD (default 0.3) — lower = more matches, higher = stricter
• The .slice(0, 10) limits to 10 calls — increase if your sales cycles involve more calls

Customization Ideas

Replace Gmail with Slack

Delete the Send Brief node and add a Slack node instead. Use $json.html for rich formatting or $json.plainText for plain text. Send to a dedicated #implementation-handoffs channel.

Add a Notion/Confluence Page

After the Format Email node, add a Notion or Confluence node to create a persistent implementation brief page instead of (or in addition to) email.

Auto-Create a Project

Add nodes after the brief to create a Jira epic, Asana project, or Monday.com board with tasks based on the identified gaps.

Different Intake Checklists by Deal Type

Add a Switch node after Salesforce that routes to different Prepare LLM Context nodes based on sf.Type or sf.RecordTypeId — each with a different intake checklist for different product lines or deal sizes.

Troubleshooting

Problem

Fix

"No matching calls found"	Check that the Gong workspace ID is correct. The matching may be too strict — lower MATCH_THRESHOLD or add more terms to strip from stripSuffixes.
Salesforce field is empty but data exists	The custom field API name in your SF org may differ. Check the actual API name in Salesforce → Setup → Object Manager → Opportunity → Fields.
Gong pagination never ends	Check the If cursor exists node — the cursor check must be notEmpty. If Gong returns empty cursor, the loop should exit to Output All Calls.
LLM response is cut off	Increase maxTokens in the Generate Brief node (default: 4000). GPT-4.1 supports up to 32K output tokens.
PDF text not included	Make sure the uploaded file is a text-based PDF (not a scanned image). The extractFromFile node only works with text PDFs.
Brief quality is poor	Switch from gpt-4.1-nano to gpt-4o or gpt-4.1 in the Generate Brief node for better analysis.