Step 2: Connect Google Sheets

These nodes read lead data and write icebreakers back to your sheet.

1. Open Retrieve Lead Rows and select the spreadsheet and sheet to read from.
2. Set the Document ID in Retrieve Lead Rows to the correct Google Sheets file and choose the Sheet Name.
3. Open Update Sheet Records and set Operation to update.
4. Map the update fields in Update Sheet Records using the existing expressions:
   • id → ={{ $('Retrieve Lead Rows').item.json.id }}
   • Icebreaker → ={{ $json.output.icebreaker }}
   • company_name_cleanup → ={{ $('Normalize Company Name').item.json.message.content.CompanyNameCleanUp }}
5. Ensure Matching Columns includes id so the update targets the correct row.
6. Credential Required: Connect your Google Sheets credentials in both Retrieve Lead Rows and Update Sheet Records.

Step 3: Filter and Batch Leads

This step ensures only leads without an icebreaker are processed and handles them in batches.

1. Configure Validate Unprocessed with the condition:
   • Left Value → ={{ $json.Icebreaker }}
   • Operation → equals
   • Right Value → no data
2. Open Batch Lead Iterator and set Batch Size to 25.
3. Verify the flow: Retrieve Lead Rows → Validate Unprocessed → Batch Lead Iterator.
4. Confirm the loopback path: Update Sheet Records → Batch Lead Iterator to process the next batch.

Step 4: Set Up AI Research and Icebreaker Generation

These nodes normalize company names, research prospects, and craft the final icebreaker line.

1. In Normalize Company Name, keep JSON Output enabled and ensure it uses your OpenAI model settings.
2. Normalize Company Name already references the organization name via {{ $('Retrieve Lead Rows').item.json.organization_name }}; verify it matches your sheet column name.
3. Connect Web Research Model as the language model for Run Company Research.
4. In Run Company Research, confirm the human message template includes:
   • {{ $('Retrieve Lead Rows').item.json.first_name }} {{ $('Retrieve Lead Rows').item.json.last_name }} {{ $('Retrieve Lead Rows').item.json.title }} {{ $('Retrieve Lead Rows').item.json.organization_name }} at {{ $now.toFormat('dd.MM.yyyy') }}
5. Connect Compose Icebreaker LLM as the language model for Draft Icebreaker Line.
6. Attach Extract Icebreaker JSON as the output parser for Draft Icebreaker Line (credentials are added to Draft Icebreaker Line, not the parser).
7. Credential Required: Connect your OpenAI credentials in Normalize Company Name (already configured as openAiApi).
8. Credential Required: Connect your OpenRouter credentials in Web Research Model and Compose Icebreaker LLM.

Step 5: Configure Output Updates

This step writes the generated icebreaker back to your sheet for each processed lead.

1. In Update Sheet Records, confirm Columns are mapped to the expressions already provided.
2. Double-check the Matching Columns includes id to update the correct row.
3. Ensure the execution flow matches: Run Company Research → Draft Icebreaker Line → Update Sheet Records.

Step 6: Test and Activate Your Workflow

Validate the end-to-end flow before enabling it for routine use.

1. Click Execute Workflow from Manual Launch Trigger to run a manual test.
2. Confirm that Validate Unprocessed only passes rows where Icebreaker equals no data.
3. Check that Draft Icebreaker Line outputs a valid JSON object and Update Sheet Records writes it to the sheet.
4. When satisfied, toggle the workflow to Active for production use.