Skip to content
‹ Back to Agent Factory
REVENUE_GEN

Outbound execution support

Drafts the email, drafts the LinkedIn message, drafts the follow-up, categorizes the reply, logs the touch, and hands off the qualified ones. Your reps stop typing and start talking.

Overview

The job of an outbound rep used to be conversations. Somewhere along the way it became typing. Drafting first-touch emails, drafting follow-ups, copying the message into LinkedIn, logging the touch in the CRM, categorizing the reply, prepping the handoff, by the time a rep finishes the admin work on twenty accounts, the day is over and the conversations didn't happen.

This agent runs the keyboard work. It drafts the first-touch email and the LinkedIn variant in the rep's voice, generates the follow-up cadence, reads incoming replies and categorizes them, interested, not interested, wrong person, out of office, ghost, logs every touch and reply to the CRM, and preps the handoff brief when a prospect is ready for a call. The rep does the human work: judgment, conversations, closing. The agent does everything else.

Capabilities

  • Drafts first-touch emails and LinkedIn messages in the assigned rep's voice, anchored to the account's research brief and the prospect's specific signals
  • Generates follow-up sequences that adapt based on prospect behavior, opens, clicks, replies, silence, instead of running a fixed cadence regardless of signal
  • Categorizes incoming replies into actionable buckets (interested, objection, not now, wrong person, OOO, soft pass, hard pass) and routes each accordingly
  • Logs every touch, sent, opened, clicked, replied, to the CRM with full context, so the activity record stays clean without rep effort
  • Preps the handoff to an AE when a prospect signals interest, with a brief that includes the full thread, the prospect's stated context, and the recommended call opener
  • Identifies pattern problems across the rep's outbound, subject lines tanking open rates, follow-up #2 killing reply rates, and surfaces them so the rep can adjust

Example Output

Trigger

SDR Marcus Bell needs to send first-touch outbound to 47 accounts on his Vanta Bio territory list this morning. Account briefs and personalization lines from the Account & Personalization agent are already attached.

Agent draft #1, Daniel Park, Director of Engineering at Vanta Bio:

Subject: Research velocity vs. platform reliability

Hi Daniel,

Saw your repost of Maya's research-velocity-vs-platform-reliability piece, the work you described in your hiring post is exactly the gap most Series B biotechs are sitting on right now.

A few of the platform leads I work with at peer biotechs have been solving this two ways: tightening the lab data → ML training loop, and standardizing the platform reliability layer before headcount scales. Worth 20 minutes if you want to compare notes on what's working and what isn't.

Open to a short call next week?

Marcus

LinkedIn variant (shorter, casual register):

Daniel, caught your repost of Maya's piece on research velocity vs. platform reliability. Working with a few platform leads at peer Series B biotechs solving the exact gap. Worth comparing notes on a 20-minute call?

Reply categorization (12 hours later):

Daniel replied: "Send me a calendar link, Tuesday or Wednesday next week works."

Category: Interested, high intent, ready for booking

Action taken: Handed off to Booking & Follow-Up agent. Marcus notified in Slack with handoff brief.

Pattern surfaced (across the week's outbound):

Marcus's outbound this week: 47 sent, 38% open rate, 11% reply rate, 4% positive reply rate.

Pattern flag: Follow-up #2 in your current cadence is dropping reply rates by 60% vs. follow-up #1. Suggest A/B testing two new variants, drafted and ready for your review.

Agent Workflow

Step 01

Receive the outbound queue

The agent receives the day's outbound queue: target accounts, prospect contacts, account briefs, and personalization lines (from the Account & Personalization agent or a connected list source). Each prospect enters the queue with full context already attached.

Step 02

Draft the first-touch message

The agent writes the first-touch email and a parallel LinkedIn variant. Both are anchored to the prospect's specific signal, recent activity, trigger event, stated context, and written in the assigned rep's voice based on prior outbound. No template smell, no mail-merge tells.

Step 03

Send or queue for rep approval

Based on the rep's config, drafts either send automatically on a schedule or queue for one-click rep review. Most teams start with review-required for the first two weeks, then move to auto-send once draft quality is trusted.

Step 04

Track the activity signal

The agent monitors opens, clicks, link engagement, and reply behavior on every sent touch. The signal feeds the next-step decision, whether to follow up, when to follow up, and what angle to use, instead of running a fixed cadence regardless of what the prospect is doing.

Step 05

Generate adaptive follow-ups

When a follow-up is due, the agent drafts the next message based on what's happened so far. A prospect who opened three times but didn't reply gets a different follow-up than a prospect who never opened. The cadence stays human-paced and stops when the signal says it should.

Step 06

Categorize incoming replies

Every inbound reply gets read and categorized: interested, objection, not now, wrong person, out of office, soft pass, hard pass. Each category triggers a specific downstream action, handoff, nurture queue, suppression, polite acknowledgment.

Step 07

Log every touch and reply

Every sent touch, every open, every click, every reply, every category decision writes to the CRM in real time. The activity record stays clean and complete without the rep typing a word into Salesforce. RevOps gets the data they've never been able to get.

Step 08

Surface patterns and prep handoffs

For interested replies, the agent preps a handoff brief, full thread, prospect context, recommended call opener, and routes to the assigned AE. Across the week, the agent surfaces patterns the rep can't see in their own activity: which subject lines are tanking, which follow-up step is leaking, which angle is converting.

Go live in days

AI for work, that just works.