OutReach AI
Multi-Channel Outbound
With Per-Prospect Tone
Generic templates ignore that the same pitch needs three different tones for three different prospects. OutReach AI generates them, predicts open and click rates before sending, sequences across email, SMS, and LinkedIn, and routes everything through approval gates.
What It Does
Five capabilities. One outbound surface.
Replaces a stack of single-purpose tools (email sequencer + tone tester + send-time picker + multi-channel scheduler + compliance approval) with one engine that knows how the pieces interact.
Three tone variants per prospect
For every prospect, OutReach AI generates the same outbound in three voices: formal, conversational, and direct. The right tone is picked based on prior interaction signals (LinkedIn writing style, prospect role, prior reply patterns) — or you pick manually before send.
Pre-send open + click prediction
Each draft gets an estimated open rate and click rate before it leaves the queue. Predictions are based on subject-line patterns, length, time-of-day, and historical reply data per channel. If a draft scores below your floor, the engine surfaces a rewrite option.
Multi-channel sequencing
A campaign isn't one channel. OutReach AI orchestrates a sequence across email, SMS, and LinkedIn — each step keyed off the previous step's response signal. If email step 1 was opened but not replied to, step 2 might switch to LinkedIn. If SMS step 3 was answered, the sequence pauses.
Send-time optimization
Per-recipient send-time learning. Once a prospect has interacted with two or three previous messages, OutReach AI knows their open patterns (Tuesday 10am vs. Thursday 4pm vs. weekends) and shifts later sends to match. New prospects get sent at channel-default optimal times.
Approval gates on every channel
SMS and email both have compliance and brand-voice failure modes. Every outbound message hits a HITL gate first — the operator reviews the draft, the predicted scores, the channel, and the sequence position, then approves, edits, or kills the send. Zoho Mail and SMS deliver the approved messages.
Zoho Mail + SMS integration
All deliveries flow through your Zoho ONE stack — your sending domain, your numbers, your audit trail. No third-party SMTP, no separate SMS provider to wire up. The integration handles bounces, opt-outs, and DKIM correctly out of the box.
How It Works
Generate → Score → Sequence → Gate → Deliver.
Per-prospect generation
Pull prospect context (role, company, prior interactions, LinkedIn writing style) and generate three tone variants of the outbound. Each variant gets its own subject line, body, and inferred channel preference. Drafts are stored — nothing sends yet.
Open + click prediction
Each draft gets predicted open and click rates. Predictions use subject-line patterns, message length, calendar position, sender domain reputation, and the prospect's historical engagement with prior messages. Below-threshold drafts surface a rewrite suggestion before they queue up.
Sequence orchestration
A campaign defines steps as (channel, delay, branch-on-signal). After step 1, the engine waits for the configured signal window (open, reply, click) and either advances, branches to a different channel, or pauses. Sequences fan out per prospect — one prospect's path doesn't constrain another's.
HITL approval gate
Every outbound message — email, SMS, LinkedIn DM — hits the approval queue. Operators see the draft, predicted scores, and the sequence context. Approve, edit, or kill. Approved drafts queue for delivery; killed drafts are logged with a reason for sequence learning.
Send-time + delivery
Approved messages route through the right Zoho ONE channel at the per-prospect optimal send time. Bounces, opt-outs, and replies feed back into the sequence for the next step. Every action is logged for audit and for the open/click prediction model.
Built honestly
14 source modules, 10 test files, deployed on port 3160 — and three things we haven't shipped yet
OutReach AI is in early access today: deployed as outreach-ai-api, the campaign and approval flow runs in production, content-package dispatch + delivery-attempt outbox patterns are wired. Three gaps remain on the honest punch list.
1. SuperTokens auth is not yet wired (audit P0 #4). Internal use is fine; before we route customer traffic through the API, this lands. Tracked in epic #44.
2. Branch-coverage measurement on a clean CI run is still pending (audit P0 #3, rescoped 2026-04-25). The audit's original "2.5% test ratio" framing was based on a wrong file-count proxy; the actual file count is 10 tests / 14 source files. What we still owe the audit is a measured branch-coverage number and, if it's below 30%, a lift to that bar.
3. The per-product Grafana dashboard is not yet provisioned (audit P2 #11). The /metrics endpoint just shipped (PR #40 wired @mos/observability); Prometheus scrape config follows next. The customer-facing operational view is the milestone after that.
We'd rather tell you what's not done than rebrand work-in-progress as polish.
Stop sending the same message in three different tones manually.
Let the engine pick the right one.
OutReach AI is in early access. Customers running their first sequences inside the gated approval flow.
Built & Verified
Subscription
$49/mo - $299/mo