AI Case Study

Claude + Gmail + Drive SOPs Drafting Every Support Reply

Customer SupportVanessa
  • “That would lower the hours I pay my VA from three a day to one — that would be thousands of dollars per month.” — Vanessa

The problem

Vanessa's VA spends about 3 hours a day sorting the inbox and drafting replies; getting that to 1 hour a day would save thousands of dollars per month. Her weekend assistant works only Saturdays and Sundays, so infrequent work plus "a lot of random things to remember" made support error-prone, and her process knowledge lived in scattered Loom tutorials and a clunky PhraseExpress snippet database that's hard to share between people.

The AI solution

Only a week and a half into using Claude, Vanessa created a Claude Project and told it: "I want you to be able to answer emails with really accurate drafts and also suggest admin next steps — what can I give you?" Claude specified a folder structure for Google Drive; she connected Drive (after learning the hard way that uploading documents directly into Claude burns through the usage quota — Drive reads don't count the same way). Into a "support knowledge" folder went every piece of documentation: pricing and promo docs, transcripts of the Loom tutorials her assistant made for the weekend assistant, a follow-up-tasks folder (e.g. what to do after sending a refund email), and the PhraseExpress database of replies to hundreds of emails.

Crucially, she had Claude read all of it and reorganize it into a master document optimized for its own retrieval. Claude realized it didn't need 10 separate documents for 10 courses that share the same refund policy — just one refund section in a master doc — and it split "finalized processes" from "email responses" on its own. The reorganization also surfaced that many 2.5-year-old tutorials were outdated, forcing a useful refresh.

In use, she or her assistant tells Claude "check the inbox" — it's connected to Gmail via the Google Workspace connector. Claude sorts emails into categories (delete — e.g. "thank you, I love you" messages, which it tags rather than deletes — spam, high priority, low priority), drafts replies to everything that needs one, and produces a second draft listing the next admin steps for emails that have follow-up actions, so nobody has to remember them.

Tools & costs

Claude Gmail (Google Workspace connector) Google Drive Loom transcripts PhraseExpress snippet database (source material) Kit

Results

Target: cut VA inbox time from 3 hours/day to 1 hour/day — "thousands of dollars per month." Early tests produced fine drafts; nuanced cases need more documentation, which she's adding by working the inbox herself for a week. Side benefit: the knowledge audit caught years of outdated SOPs.

How to copy it

  1. Create a Claude Project and ask Claude what documentation it needs to draft your support emails; let it design the Drive folder structure.
  2. Connect Google Drive (don't upload docs directly — you'll hit usage limits) and fill the folders with pricing docs, SOPs, Loom transcripts, and your saved reply snippets.
  3. Have Claude read everything and consolidate it into one master document organized for its own retrieval, deduplicating shared policies.
  4. Connect Gmail via the Google Workspace connector and prompt "check the inbox" — have it sort by category, tag (not delete), and draft replies.
  5. Ask for a second draft per email listing next admin steps, and add documentation whenever a nuanced case gets a weak draft.

Source

AI Hangout — July 2026 (recording) · watch this segment on YouTube (18:30) →