I tried to let Gemini write my emails for a week — here is why I had to intervene

Our inboxes used to be for talking to people, but now they're basically just a mountain of digital debt.

Most of us spend hours every single day trying to stay afloat in a sea of emails.

Every time I see a "just circling back" or a "touching base," it feels like my mental health is taking a hit, and I've always wanted a way to offload this nonsense.

It turns out a lot of people feel the same way, which explains why Gemini showed up in Google Workspace in the first place.

Since I was desperate to see if it actually lived up to the hype, I came up with a five-day test plan to let the AI take over my inbox.

Related

7 unexpected ways I use Gemini beyond chat

It became a crucial part of my Android workflow

Posts By  Anu Joy

Days 1–3: When the AI feels like a superpower

The first three days felt like cheating.

Gemini was in its element handling the boring, repetitive stuff. Scheduling meetings, confirming receipts, and quick no-thanks replies were not my problem anymore.

There's something deeply satisfying about tapping the Generate button and watching a clean reply appear.

Unlike standalone chatbots that live in their own little bubbles, Gemini has context awareness. It knows who you're replying to and what's already been said in the thread.

When you turn on the Workspace setting, it can cross-check Google Drive and Calendar without you spelling everything out.

A boring coordination email is the perfect test case. A coworker emailed about an invoice they couldn't find from last month.

The usual routine is to open Drive, search for filenames, grab the PDF, and forward it along. But all I did was prompt:

Find the receipt for the Dell monitors from last March in my Drive and draft a response.

Gemini executed it perfectly, start to finish. That kind of speed matters when time is the thing you're always short on.

Day 4: Sounding like a corporate bot

Day four was my "wait a second" moment.

After you've sent enough AI-written emails, the pattern becomes evident. Gemini defaults to a tone that's professional, yes—but also polished to the point of feeling lifeless.

It loves to throw in the same polite filler lines, like "Hope this finds you well" and "Thanks for your patience," over and over. After a while, they feel fake and distant.

Losing my personal touch didn't go unnoticed. Friends and longtime colleagues started picking up on it.

Relationships are what hold a team together. They rely on sincerity and unspoken social agreements. Gemini was making conversations feel insincere, and I had to step in.

Now, I have to own some of this. I started out giving Gemini really lazy prompts, and I got back generic, lazy garbage.

I tried to fix it by being more specific. That definitely helped, but you could still see the robot accent.

And it honestly begs the question: If it takes me three minutes to come up with the perfect prompt just to get it to sound human, what's the point?

Subscribe to the newsletter for smarter AI-email use

Looking for clearer guidance on using AI in your inbox? Subscribe to the newsletter for hands-on prompt templates, tone adjustments, and practical rules of thumb about AI-written email — focused on helping you use AI without losing your voice.

Subscribe
By subscribing, you agree to receive newsletter and marketing emails, and accept Valnet’s Terms of Use and Privacy Policy. You can unsubscribe anytime.

Might as well write the email myself.

Day 5: Putting the AI on junior intern duty

I didn't fire Gemini by day five, but I did put it on junior intern duty.

Gemini is a great partner for pulling things together and structuring them, but it's not reliable for the final output. That's where you see its limits.

Gemini struggles with gut feelings and the emotional intelligence that sensitive conversations require. I

t can't fully capture your personal voice, or really tell when it's okay to crack a joke. That's why I've come up with some ground rules for AI in email:

• Never "insert and send": Always "insert, edit, and then send."
• No AI for apologies: If I make a mistake, I write the apology. AI apologies sound sociopathic.
• No AI for negotiations: Nuance is lost.
• Use it for synthesis: Summarizing long threads is its best use case, provided I verify the actionable claims.

The danger of becoming a passive observer

When I finally had to write an email by myself, I froze. After so much autopilot, even a simple email felt exhausting.

We worry about AI becoming sentient, but the bigger risk might be us becoming lazy. Passing off the thinking while we scroll TikTok is mental decay.