AI-assisted workflow to process editorial notes into Airtable

Questions
, , ,
1

:bullseye: What is your goal?

I am building an AI-assisted workflow to analyze and process research notes into structured data in Airtable for easy scanning.

:thinking: What is the problem & what have you tried?

I have built several scenarios to automate an editorial workflow that processes research notes and uses AI to summarize and feed responses into Airtable. All scenarios so far have failed to produce usable text in Airtable, and I have abandoned all but the most recent build. The trigger is when a new .doc file uploads in a Google Drive folder. Each .doc file contains research for a given article topic (notes, links, thoughts, etc). The scenario should pass that file to Gemini (2.5 Flash) for analysis, and Gemini should output summarized data into Airtable. What should be a fairly simple automation has been a nightmare of endless errors. Sometimes they are rate limit errors, so I upgraded my Google API to a paid billing tier. Everything else in the automation chain is free tier for now. The biggest error seems to be that Gemini is reading text from the original document as data. I tried inserting filters and/or modules to convert to .txt before analysis, but no luck. Gemini can handle summarizing these documents when I feed it an individual .doc file directly, but once Gemini is in the middle of a Make scanario, it can’t process the information and get it into Airtable. The endpoint should be an Airtable table with each document/article file represented by one row with multiple columns for the analyzed data. Instead, it just outputs irrelevant information, or none at all. I have not even added the Airtable module in the most recent build (see attached) because it seems like the information is not being processed. I have added Airtable in previous Scenarios, along with much more elaborate input prompts, but that would seem pointless in this one if the output is not useful. Frustrated and about to give it up.

Note: Not a coder here. Please keep responses easy to understand.

Thanks!!

:camera_with_flash: Screenshots (scenario flow, module settings, errors)

Make_Output.png
Make_Output.png3762×1854 546 KB

1 Like
3

Hey Steve,

how are you passing the file at the moment? Did you try using the upload a file module?

4

I started to add an Upload a File mod between the Google Drive-Download a File and Gemini-Generate a Response mods, but I bailed because I wasn’t sure how to configure it. Never ran a test on it. Below is the latest module sequence recommended by one of several AI models I have consulted.

Does it sound right? After the first two mods, I get confused about how to configure and map, so right now I have only the first two GD mods and the Gemini mod.

Google Drive → Watch files in a folder
Google Drive → Download a file
Google Drive → Upload a file ← CONVERSION HAPPENS HERE
Google Docs → Get Content of a Document
Google Gemini AI → Generate a response

GPT’s analysis based on a screenshot of the output:
4. The core problem (in one sentence) You are passing a binary file into a text-only AI field without converting it to plain text first.

Thanks for your help!

1 Like