Generate financial risk reports from ElevenLabs interviews using OpenAI and Gmail

This workflow automates the process of receiving a post-call audio file and transcription from ElevenLabs, processing them, and generating a financial risk report.

Key Advantages

1. ✅ End-to-End Automation

The workflow fully automates the process from raw input (audio/transcript) to final delivery (email report), eliminating manual intervention.

2. ✅ AI-Powered Decision Making

It leverages language models to:

• Analyze qualitative interview responses
• Convert them into quantitative scores
• Produce consistent and objective evaluations

3. ✅ Structured Data Extraction

Automatically extracts critical business information, reducing human error and ensuring standardized outputs.

4. ✅ Scalability

The webhook-based architecture allows the system to handle large volumes of interviews in parallel without additional effort.

5. ✅ Modular & Extensible Design

Each step (audio processing, extraction, scoring, reporting) is modular, making it easy to:

• Replace models
• Add new analysis layers
• Integrate additional services

6. ✅ Professional Output Generation

Generates clean, ready-to-send HTML reports compatible with email clients, improving communication with stakeholders.

7. ✅ Data Traceability & Storage

Audio files are stored in Google Drive, ensuring:

• Auditability
• Easy retrieval of original data

8. ✅ Consistency & Standardization

The evaluation logic ensures that all interviews are assessed using the same criteria, reducing subjective bias.

How it works

1. Receiving and Routing Data: The workflow starts with a Webhook that listens for incoming data from ElevenLabs. A Switch node then routes the data based on the body.type field.

   • Post Call Audio: If the type is post_call_audio, the workflow processes the audio.
   • Post Call Transcription: If the type is post_call_transcription, the workflow processes the transcription.

2. Audio Processing Path: For an audio file, a Code node extracts the Base64 audio data and the conversation_id from the webhook payload. It converts the Base64 string into a binary audio buffer (MP3). This binary data is then passed to a Google Drive node, which uploads the file to a specified folder (the user's root folder).

3. Transcription Processing Path: For a transcription, a Set node extracts the transcript array from the payload. A subsequent Code node processes this array, combining all messages from the conversation into a single, readable full text string, prefixed by the speaker's role.

4. Data Enrichment and Analysis: The full transcript text is then used by two nodes in parallel:

   • Information Extractor: This LangChain node uses an OpenAI model (gpt-5-mini) to extract structured data from the text, specifically the company_name, the CEO's name, the address, and the vat_number.
   • Calculate Rating: This LangChain node uses another OpenAI model to perform a quantitative evaluation. It follows a provided system prompt to assign a numerical score, a final verdict (POSITIVE/NEUTRAL/NEGATIVE), and a reason based on the interviewee's responses. Its output is parsed by a Structured Output Parser to ensure it is valid JSON.

5. Report Generation and Delivery: The outputs from the Information Extractor and Calculate Rating nodes are merged into a single data object. This object is passed to the Financial Report Generator, a final LangChain node that acts as a professional analyst. Using the merged data (company details, score, verdict, etc.), it generates a polished, formatted HTML email body. Finally, a Gmail node sends this HTML report as an email to the specified recipient.

Set up steps

1. Configure Credentials:

   • OpenAI: Set up an OpenAI API credential for the three language model nodes. Ensure it has access to the gpt-5-mini model.
   • Google Drive: Configure OAuth2 credentials for the "Upload audio" node to allow file uploads.
   • Gmail: Set up OAuth2 credentials for the "Send report" node.

2. Configure Webhook:

   • Note the webhook ID and path. This URL must be configured in ElevenLabs to send post-call data to this n8n instance.

3. Update Node Parameters:

   • Google Drive: Modify the "Upload audio" node if the target folder (folderId) is not the root.
   • Information Extractor: The extraction attributes (company, name, address, VAT) are pre-configured. No changes are needed unless the target data fields change.
   • Gmail: Update the Gmail node with the recipient email address ([email protected]) and verify the email subject line formatting.

4. Activate Workflow: Once all credentials and parameters are set, toggle the workflow from active: false to active: true in the n8n editor to start listening for webhook calls.

👉 Subscribe to my new YouTube channel. Here I’ll share videos and Shorts with practical tutorials and FREE templates for n8n.

Need help customizing?

Contact me for consulting and support or add me on Linkedin.