CarcMail Product Updates — June 2026: AI Rewrite Engine, Sequence Branching, and Reply Intelligence v2

What Shipped in June 2026

This has been one of the largest shipping months in CarcMail’s history. Four major features landed in production, two long-standing bugs were fixed, and three quality-of-life improvements rolled out to all accounts. Here’s everything.

---

1. Rebuilt AI Rewrite Engine with Tone Control

The original AI rewrite feature rewrote flagged emails in a single “safer” style — lower spam risk, but sometimes at the cost of personality. The June rewrite overhauled this entirely.

What changed:

• Tone sliders — Choose between Formal, Conversational, Direct, and Friendly. The AI rewrite now preserves the style you set in your AI Identity profile rather than defaulting to neutral.
• Preserved personalisation — First name, company name, and custom fields no longer get genericised during rewrite. The rewrite targets risk patterns (spam trigger words, excessive punctuation, broken link structures) without touching personalisation tokens.
• Rewrite explanation — Each rewrite now includes a one-line explanation of what was changed and why. You can see “Removed urgency trigger phrase in subject line” rather than getting a black-box result.
• Side-by-side diff view — Before/after comparison is now shown inline in the campaign editor. Accept, reject, or edit the suggestion directly.

The new engine uses claude-sonnet-4-6 as primary with gpt-4o-mini as fallback. Average rewrite latency dropped from 4.2s to 1.8s.

---

2. Multi-Branch Follow-Up Sequences

Previously, CarcMail sequences were linear: Email 1 → wait 3 days → Email 2 → wait 5 days → Email 3. No branching, no conditions.

June 2026 introduces conditional branches based on reply classification:

How it works:

• If a lead replies and is classified as Interested → branch to a booking-focused follow-up (shorter, specific CTA)
• If classified as Not Now → branch to a long-tail sequence (bi-weekly, lower pressure)
• If classified as Unsubscribe → automatically suppressed from all future sends
• If no reply after the final email → move to a re-engagement branch or mark as closed

You define the branch conditions when building the sequence. CarcMail’s Reply Intelligence engine runs the classification automatically — no manual sorting required.

Branching supports up to 4 conditions per step and up to 6 total branches per sequence. Available on Growth and Scale plans.

---

3. Reply Intelligence v2

The original reply classifier covered four categories: Interested, Not Now, Unsubscribe, and Other. Version 2 expands this to nine:

Classification	What It Means
Interested	Positive signal, wants to continue
Meeting Booked	Confirmed calendar appointment
Not Now	Timing objection, re-engage later
Referral	Points to another decision-maker
Out of Office	Auto-reply, not a real response
Negative	Hard rejection, do not contact again
Unsubscribe	Explicit removal request
Question	Has a specific question, needs answer
Other	Uncategorised, review manually

What’s new:

• Out-of-office detection — OOO replies are no longer miscounted as real responses. The lead stays in the sequence and a follow-up is sent when the OOO period ends (if a return date is detected).
• Referral extraction — When a reply contains another person’s name or email (“talk to Sarah at this address”), CarcMail extracts it and adds it to a review queue for manual addition as a lead.
• Question routing — Replies classified as Question are flagged in your inbox with a “Needs reply” badge so they don’t get buried.

The classifier now runs on claude-sonnet-4-6 with a structured JSON output schema — accuracy improved from 81% to 94% on our internal benchmark set.

---

4. Warmup Algorithm v3: 14-Day Standard Warm-Up

Domain warmup used to take 28 days for a fresh domain to reach a stable 200-email/day sending volume. June’s warmup algorithm ramp-up is now:

Day	Daily Volume
1–2	5
3–5	20
6–9	60
10–12	120
13–14	200

The new algorithm monitors reply rates, bounce rates, and spam placements in real time and adjusts the ramp accordingly. If bounce rate exceeds 2% on any day, the ramp pauses automatically and you’re notified.

The warmup pool was also expanded — warmup emails now cycle across 2,400+ real mailboxes (up from 800), which produces more natural engagement signals for ISP reputation scoring.

Important: The 14-day target assumes a clean domain (no prior send history). Domains with reputation damage from previous senders may take longer. The dashboard now shows a warmup health score (0–100) so you can see exactly where you stand.

---

Bug Fixes

• Fixed: Sequence enrollment count was double-counting leads who were enrolled then unenrolled and re-enrolled. The count now reflects active enrollments only.
• Fixed: The spam check was flagging CarcMail’s own unsubscribe link footer as a spam signal. The link is now whitelisted from the spam scoring model.

Quality-of-Life Improvements

• Lead timeline — Each lead’s profile now shows a full activity timeline: email sent, opened, clicked, replied, bounced, unsubscribed. Previously only send events were visible.
• Campaign capacity indicator — The dashboard now shows remaining daily send capacity per inbox account so you know before starting a campaign whether you have enough headroom.
• Export improvements — CSV exports now include reply classification labels and sequence branch status alongside the existing send log fields.

---

What’s Next

July 2026 roadmap (subject to change):

• A/B testing for subject lines — Split test two subject line variants across a campaign with automatic winner selection after 200 sends
• Apollo enrichment in-app — Search and enrich leads without leaving CarcMail
• Webhook support — Push send, open, reply, and bounce events to external systems via webhook