Controlled automation: why InboxOji drafts, and you decide

“Let AI handle your inbox” is an easy thing to promise and a dangerous thing to ship. The moment software starts sending email on your behalf, one confident-but-wrong reply can cost you a client, a deal, or a lot of trust. So when we built InboxOji, we drew a hard line — and that line is the whole product.

The problem with full autopilot

Most “AI email” tools optimize for a demo: watch it read a message and fire back a reply, hands-free. It looks magical. But your inbox isn’t a demo. It’s contracts, refunds, apologies, prices, promises. The cost of a bad automated send isn’t “redo it” — it’s a real conversation with a real person who now thinks you said something you didn’t.

Full autopilot asks you to trust a model with your reputation on every message. We don’t think that trade is worth it for most of what lands in your inbox.

The line we drew: draft, don’t send

By default, InboxOji drafts. It reads the emails you’ve told it to watch, writes a reply in your voice, and leaves it waiting for you — the same as if a sharp assistant had it ready when you opened your laptop. You read, tweak if you want, and hit send. Nothing leaves your account because a model decided it should.

The AI does the typing. You keep the judgment.

That sounds modest. In practice it’s the difference between a tool you can actually run your business on and one you have to babysit.

How control actually works

You decide what InboxOji touches with one simple idea: a rule. A rule is a situation you handle often (“new booking request,” “invoice question”) paired with the reply you’d send. InboxOji reads each email, drafts only when it matches a rule you created, and leaves everything else alone. If none of your rules fit, it drafts nothing — no forcing a square reply into a round email.

It writes in your voice, not a generic “Dear valued customer” one — you set the tone once and refine it as you go. And it never invents facts: your booking link, address, and prices come from values you save, dropped in exactly right every time.

When sending itself is fine

Some replies really are safe to send automatically — the truly repetitive, low-stakes ones. For those, you can turn on auto-send per rule, and only after you’ve tested the rule and seen it get the answer right. It’s opt-in, scoped, and reversible — the opposite of “automate everything and hope.”

Time back, not control given up

The goal was never to remove you from your inbox. It was to remove the retyping — the same answer to the same question for the hundredth time — while leaving every real decision with you. That’s controlled automation: quietly handled in the background, always under your name, always your call.