Playbook

AI replies that sound like you — not like ChatGPT

Why we fine-tune per-user instead of using prompt engineering, and how the Autocloz reply-drafter learns your voice from your last 200 sent replies.

2 Apr 2026 7 min readBy Autocloz Editorial, AI team

Prompt engineering hits a ceiling

You can prompt your way to a polite, generic reply. You cannot prompt your way to "this sounds like Vikram, not like ChatGPT."

Per-user fine-tuning

Autocloz fine-tunes a small adapter on your last 200 sent replies. The base model stays shared; the adapter is per-user. Drafts inherit your sentence length, your typical openers, your specific verbal tics.

Always human-in-the-loop

The AI never auto-sends. Every draft requires a click. This is by design — the speed gain comes from drafting, not from removing the operator.

Keep reading

Playbook

Why we don't resell carrier minutes — and why you shouldn't pay anyone who does

Aircall and JustCall mark up Twilio rates 3–5×. We pass Telnyx and DIDLogic rates through verbatim. Here's the math, the architecture, and the hidden cost you're paying.

Multi-channel

We replaced Twilio Studio with a visual IVR — here's why the simulator matters

Most IVR builders are paint-on-glass. The simulator runs different code than production, so flows that pass the simulator misbehave on real calls. We rebuilt it as one state machine.

Deliverability

Cold email deliverability in 2026 — what actually changed

Gmail and Yahoo's 2024 sender requirements moved the floor. DMARC enforcement, one-click unsubscribe, and bounce-rate gates are now table stakes. Here is the field-tested checklist.

Free to start

Stop reading. Start sending.

Every tactic in this article is implemented behind the Autocloz dashboard.

Start free Talk to sales