Speed up pitch deck reviews by turning uploads in Airtable into clear text, a structured report, and a chat tool your team can use to ask questions. Ideal for marketing and growth teams that collect decks and need fast insights without reading every slide.

Here is how it runs end to end. New or pending rows in Airtable kick off a run. The file is fetched, converted from PDF to page images with a simple HTTP service, and unzipped. Pages are sorted and resized. An AI vision model reads each page and writes clean markdown. All pages are combined, then a report is created and the record in Airtable is updated. The text is split into chunks, turned into embeddings, and stored in a Qdrant vector database. A chat webhook lets users ask questions. The agent searches the vector store to answer with context from the right deck.

To set this up you need access to Airtable, OpenAI, Qdrant, and a PDF to image service. You can use a public PDF tool or host your own for privacy. Expect reviews to drop from hours to minutes, with cleaner data in your base. Great for teams that compare many decks or need fast answers without opening files.

• Airtable trigger and search to find pending decks and start separate executions
• HTTP download of the PDF file and conversion to page images through a PDF to image service
• Zip extraction, page sorting by file name, and image resizing for better AI reading
• Vision model transcribes each page to markdown and describes charts and visuals
• Pages combined and summarized into a report, then the Airtable row is updated
• Text is split into chunks, embedded with OpenAI, and stored in a Qdrant vector collection
• Old vectors are removed before reindexing to keep the collection clean
• Chat webhook with memory and a vector store tool to answer questions about selected companies

• Reduce manual review from 2 hours to 10 minutes per deck
• Automate up to 80 percent of data entry into Airtable
• Improve extracted data accuracy by 90 percent versus copy and paste
• Handle 10 times more pitch decks each week with the same team
• Connect Airtable, OpenAI, Qdrant, and a PDF tool in one flow
• Let teams ask questions about decks without opening the files

1. Import the template into n8n: Create a new workflow in n8n > Click the three dots menu > Select 'Import from File' > Choose the downloaded JSON file.
2. You'll need accounts with Airtable, OpenAI, Qdrant and Stirling PDF. See the Tools Required section above for links to create accounts with these services.
3. In Airtable, create a table with a Name field and a File field for the deck. Add any extra fields you want the report to fill. Note the base and table details.
4. In the n8n credentials manager, open the Airtable nodes and create a new Airtable Personal Access Token credential. Paste your token and give the credential a clear name. Map the base, table, and view fields in the Airtable nodes.
5. Open the OpenAI nodes and create a new OpenAI API Key credential. Generate an API key in your OpenAI account and paste it into n8n. Save and test by running a small sample.
6. Open the Qdrant nodes and create a new Qdrant credential with your API URL and API key. Set the collection name in the vector store nodes. Make sure the same name is used across all vector nodes.
7. Open the HTTP Request node that converts PDF to images. Set the URL to your PDF to image service. You can use the public Stirling PDF endpoint or your self hosted instance for privacy.
8. Check the Extract Zip, Sort Pages, and Resize Images nodes to confirm they run in order. Keep the sort by file name so pages stay in the right sequence.
9. Confirm the Transcribe to Markdown and Generate Report nodes point to your OpenAI credential. Review the prompts if you want a different report style.
10. Enable the Airtable trigger if you want automatic runs. For testing, click Test workflow on the manual trigger to process pending rows.
11. Open the chat trigger node, copy the test or production URL, and send a sample message with a company name. Verify the reply includes context from the right deck.
12. Validate end to end by uploading a sample deck to Airtable. Check that the Airtable row is updated, vectors appear in Qdrant, and the chat responds with correct answers.
13. Troubleshoot common issues: ensure the file is a PDF, confirm API keys are active, check the vector collection name, verify the PDF service is reachable, and match Airtable field names exactly.