I skipped the AI inbox everyone's using

I'm Jeremy Hurst, VP of Autonomous GTM at Swan. I left a $450K VP Sales role to take a job with a made-up title. I did it because I got a glimpse of the 100x seller - it stopped being a thought experiment and became the actual playbook. This is a field journal for GTM and growth leaders who can feel the model changing under their feet and prefer changing with it to being left behind.

I'm finally trying out an AI-native email inbox, but you might be surprised to know which one..

I've heard the hype for a while. Roi Ravhon swears by Superhuman. Amos Bar-Joseph seems to be moonlighting for Shortwave.

So I looked at both. Read the reviews. Watched more youtube videos than I'd like to admit.

I was close to picking one.

Then I stopped & asked myself something simpler: do I really need another tool?

Turns out what I'm looking for may be staring me in the face already.

So instead of buying a tool, I built a skill. I told Swan AI to read an email, pull the full account context - every call, every prior touchpoint - and draft the reply I would've written with an hour + perfect memory.

Not generic. Mine.

The first few replies weren't very good.

Too long & verbose. Stuffed with AI slop & grandiose pronouncements I'd never actually say.

So I dove into the unglamorous part. With every draft, I deleted most of it & rewrote it by hand. Day by day, it got better.

Then I built the layer that mattered. Every finalized email became a reference example - tagged, indexed, added to a growing library. Not just examples of what TO do, but also a library of what NOT to do. Turns out the bad examples taught Swan as much as the good ones.

That's when it started compounding.

Then an RFP landed a couple weeks ago. Same skill. When we finished, I created a sub-skill just for RFPs and dropped the response in as an example. So the next one will pattern-match to the last.

Same for security questionnaires, vendor onboarding docs, and so on.

Here's what's up next: a daily trigger. Wake up, everything from yesterday already drafted, queued for approval. Or maybe I run it twice a day - midnight and noon - so morning emails get replies before I sign off in the evening.

Jury's still out on whether this scales the way a dedicated tool would, or whether I end up testing one of those inboxes anyway.

But the early progress feels compelling..

But importantly, this isn't really about email.

It's another use case built on the same foundation as every other part of our GTM machine. One context layer, getting smarter every time I use it.

What's the saying?

"Compounding is the eighth wonder of the world." I think that guy Einstein was on to something..

getswan.com

THE SIGNAL

I spent weeks comparing AI tools when the real lever was always the context I fed it, one correction at a time.

THE PLAY

Build a rejection file

• Take the next AI draft you reject and don't just delete it, save it
• Write one line next to it on why it was wrong: too generic, wrong tone, missed context
• Start a running file of rejected drafts with your notes attached
• Feed that file back in alongside your good examples as what not to do
• Check your rejection rate again in 30 days

SIT WITH THIS

If you're still comparing AI tools instead of the context you're feeding them, you're solving the wrong problem.

COMMUNITY NOTES

[Read more.

](https://lnkd.in/p/eYqeReHB?utm_source=jeremy-hurst.beehiiv.com&utm_medium=referral&utm_campaign=i-skipped-the-ai-inbox-everyone-s-using)

Stop and ask yourself. Do you really need another tool?

-Jeremy