InboxTriage

The Client

A solo founder running a 7 figure B2B SaaS company with a team of 6. No executive assistant. No inbox zero discipline. Just a Gmail account receiving 150+ emails a day from customers, investors, partners, and the usual noise.

The Problem

The founder was spending the first 2 hours of every workday on email. Not replying strategically, just triaging. Deciding what mattered, what could wait, what needed a human reply, and what was noise.

Three things made it worse:

High stakes buried in low stakes. A critical customer churn signal would land in the same inbox as a cold pitch, a newsletter, and an invoice.

Reply fatigue. Most emails needed a 2 sentence reply the founder had already written 50 times before.

Off hours anxiety. Checking email at night "just in case" had become a habit, and it was not healthy.

Off the shelf tools (Superhuman, Shortwave, SaneBox) helped at the margins but did not understand context. They could sort by sender, not by what actually mattered to this specific founder this specific week.

What They Wanted

A simple tool that would:

1. Read every incoming email and categorize it (Needs Reply, FYI, Auto Handled, Ignore).

2. Draft replies in the founder's voice for the emails that needed one.

3. Surface only the 5 to 10 emails that actually needed a human decision.

4. Stay out of the way otherwise.

How We Approached It

We spent one call going through the founder's last 200 emails with them, labeling each one as they would have manually triaged it. That became the training context.

Three decisions shaped the build:

Claude over GPT for tone matching. The founder's reply style was warm and direct. Claude matched it more naturally after we fed it 30 of their real past replies as examples.

No auto sending. Ever. Every AI draft required one click to approve. This was non negotiable for the founder, and we agreed it was the right call.

Built as a Chrome extension, not a separate app. The founder already lived in Gmail. We were not going to pull them out of it.