This is the closest thing to hiring a personal assistant without actually hiring one. You connect Claude to your Gmail once, teach it your voice with one prompt, and from then on it can go through your entire inbox, tell you what actually matters, and leave replies sitting in your drafts folder, written the way you write, ready for you to hit send. Set it up as a scheduled task and it all happens every morning before you even wake up.

Here's the full setup, start to finish, including the exact prompt from the video.

Step 1: Connect Claude to your Gmail

  1. Open the Claude app and click Customize in the top left.
  2. Under Connectors, click Add, then Browse connectors.
  3. Find the Gmail connector and press Connect. It'll ask you to sign in to your Google account and give permission. That's the part that lets Claude actually read your emails and create drafts.

One honest note: you're giving an AI read access to your inbox, so use your own judgment on which account you connect. Claude creates drafts, it doesn't send anything. You always review before anything goes out.

Step 2: Teach it your voice

Once Gmail is connected, go to Cowork, open a new chat, click New Task, paste this into the text box, and hit send. This is the prompt that locks in your voice, so the drafts sound like you and not like a chatbot.

The prompt
You are my personal email assistant, and your first job is to learn to write exactly like me.

STEP 1: LEARN MY VOICE
Go through my Gmail and read 30 to 50 emails that I SENT (not received), across different types: quick replies, longer explanations, scheduling, saying no, asking for things, and follow-ups. From those, build my voice profile:
- How I open and close emails (greetings, sign-offs, or lack of them)
- My average sentence and paragraph length
- How formal or casual I am, and with whom (boss vs friend vs stranger)
- Words and phrases I actually use, and ones I never use
- How I say no, how I ask for things, how I handle awkward messages
- My punctuation habits (dashes, exclamation points, emojis, lowercase)

STEP 2: SHOW ME
Write out the voice profile as a short style guide, then write 3 sample replies in my voice to 3 real emails currently in my inbox so I can check the match. Do not send or draft anything yet.

STEP 3: LOCK IT IN
I'll tell you what's off. Adjust until I say it's right. From then on, every reply you ever draft for me MUST follow that voice profile.

RULES
- Never send an email. You only create drafts. I review everything.
- Never invent facts, commitments, dates, or numbers. If a reply needs information you don't have, put [CHECK THIS] in the draft where I need to fill it in.
- Match the other person's language and my relationship with them. My boss and my best friend do not get the same tone.
- Keep replies as short as I would actually write them.

It'll come back with your style guide and three sample replies. Tell it what's off, "less formal," "I never use exclamation points," "shorter," until the samples read like you wrote them. That usually takes one or two rounds.

Step 3: Turn it into your morning assistant

Once your voice is locked in, open another chat and send this:

The daily instruction
Every morning at 7am, read every unread email in my inbox, prioritize them by urgency, and draft a reply to each one in my voice. Leave them in drafts so I can review before sending.

For the briefing, give me:
1. URGENT: needs a response today, with a one-line summary of each
2. IMPORTANT: needs a response this week
3. DECISIONS: emails where you need my call before drafting, with the options laid out
4. IGNORED: newsletters and noise you skipped, in one line

Then list every draft you created and who it's to.

Claude goes through the inbox, gives you the full morning briefing, tells you which emails are urgent, which are important, and which actually need a decision from you, and it uses the Gmail connection to create the drafts right in your inbox, ready to send.

The step further: make it run before you wake up

Ask Claude to turn that instruction into a scheduled task and tell it what time to run. That's it. Every morning at 7am it reads everything that came in, builds the briefing, and stages the drafts, all before you've picked up your phone. You wake up, read one summary, approve a few drafts, and your inbox is handled.

How to get the most out of it

  • Feed it more of your sent mail if the voice is off. The more real examples it reads, the closer the match gets.
  • Keep the review habit. The whole system is built so nothing sends without you. That's a feature, not a limitation. Ten seconds per draft keeps you in control.
  • Tell it your priorities once. "Anything from my boss or clients is always urgent, newsletters never are." It'll use that in every briefing.
  • Watch for [CHECK THIS]. That's the flag it drops when a reply needs a fact only you know. Fill it in before sending.

The honest part

This won't handle the emails that need real judgment, negotiations, sensitive conversations, anything where the stakes are high. It's not supposed to. It kills the 80% of inbox work that's triage and routine replies, so the only thing left for you is the 20% that actually needs your brain. That's exactly what a great assistant does.

A personal assistant who reads everything, knows what matters, and preps every reply in your voice used to cost a salary. Now it's a connector and two prompts.

Set this up tonight and let tomorrow's 7am run be the test. The first morning you wake up to a finished briefing and a drafts folder full of replies that sound like you, you'll get it.
– Anir

If this is the kind of AI thinking you want working on your own business, apply to work with me. And to catch the next breakdown, join the newsletter.