Brainforge Knowledge

MESSAGE_SEQUENCE_SYSTEM_README

8 min read

Message Sequence System - Complete Guide

Purpose: Comprehensive guide to the GTM Message Sequence Agent system
Based on: Vercel Lead Agent architecture
Status: Reference Implementation
Last Updated: 2026-01-26

🎯 What is This?

A Vercel-inspired AI agent system for generating and approving personalized, multi-step GTM message sequences across different campaign types with human-in-the-loop Slack approval.

Think: Vercel’s lead agent, but for outbound/inbound sequencing across multiple campaigns.

✨ Key Features

  • 🤖 AI-Powered Sequence Generation - Uses AI SDK to generate personalized sequences
  • 📊 Multi-Campaign Support - Event follow-up, mutual intros, cold outbound, partnerships, etc.
  • 🔄 Durable Workflows - Reliable multi-step execution with Workflow DevKit
  • 🔍 Deep Research Agent - Comprehensive prospect research with AI SDK Agent class
  • 👤 Human-in-the-Loop - Slack approval before sending (approve/edit/reject)
  • 📈 Easy to Extend - Add new campaigns in ~60 minutes

🏗️ System Architecture

Campaign Trigger
     ↓
Workflow Start
     ↓
Research Agent ──→ Prospect intel, personalization signals
     ↓
Campaign Classifier ──→ Identify campaign type, segment, persona
     ↓
Sequence Generator ──→ Generate multi-step personalized sequence
     ↓
Slack Approval ──→ Human review (approve/edit/reject)
     ↓
Schedule & Send ──→ Queue and send approved sequence

Tech Stack:

  • Next.js 16
  • Vercel AI SDK (generateObject, Agent class)
  • Workflow DevKit (durable execution)
  • Slack Bolt + Vercel adapter
  • Vercel AI Gateway
  • HubSpot (CRM integration)

📂 File Structure

gtm/agents/
├── MESSAGE_SEQUENCE_AGENT.md           # Main architecture doc
├── CAMPAIGN_TYPES_SCHEMA.md            # TypeScript schemas for all campaign types
├── SLACK_INTEGRATION_SPEC.md           # Slack integration details
├── HOW_TO_ADD_CAMPAIGNS.md             # Step-by-step guide to add campaigns
├── MESSAGE_SEQUENCE_SYSTEM_README.md   # This file
│
├── playbooks/                           # Campaign playbooks (markdown)
│   ├── event-follow-up-playbook.md     # Event follow-up sequences
│   ├── mutual-intro-playbook.md        # Mutual introduction sequences
│   ├── cold-outbound-playbook.md       # Cold prospecting sequences
│   └── [your-campaign]-playbook.md     # Add more campaigns here
│
├── memory/                              # Existing GTM knowledge base
│   ├── event-follow-up-playbook.md     # (Original file)
│   ├── bdr-tactics.md
│   ├── message-templates.md
│   └── ...
│
└── GTM_Strategy_Playbooks.xlsx         # Your original spreadsheet reference

🚀 Quick Start

1. Understand the System

Read in this order:

  1. MESSAGE_SEQUENCE_AGENT.md (10 min)
    Architecture, workflow, tech stack, key concepts

  2. CAMPAIGN_TYPES_SCHEMA.md (5 min)
    TypeScript schemas, campaign types, personas

  3. SLACK_INTEGRATION_SPEC.md (10 min)
    Slack setup, Block Kit, approval flow

  4. Pick a playbook to study:

2. Add Your First Campaign

Follow: HOW_TO_ADD_CAMPAIGNS.md

Steps (60 minutes total):

  1. Research & Planning (15 min) - Define use case, segments, personas
  2. Playbook Creation (30 min) - Write markdown playbook
  3. Schema Updates (5 min) - Add types
  4. Code Integration (10 min) - Import and configure
  5. Testing (15 min) - Validate with test triggers

Checklist:

[ ] Defined use case and segments
[ ] Created playbook markdown file
[ ] Updated campaign type enum
[ ] Created playbook import
[ ] Updated playbook registry
[ ] Created test fixture
[ ] Tested classification
[ ] Tested sequence generation
[ ] Tested Slack approval

3. Test with Your Excel Data

Your GTM_Strategy_Playbooks.xlsx file contains campaign examples. Use it as reference:

Steps:

  1. Open Excel file and review tabs (each tab = potential campaign type)
  2. For each tab:
    • Note the campaign type/use case
    • Extract segment variations
    • Copy message templates
    • Identify multi-step sequences
  3. Convert to playbook markdown using template in HOW_TO_ADD_CAMPAIGNS.md

Example Conversion:

Excel Tab: "Mutual Intro Sequence"
   ↓
Campaign Type: mutual-intro
   ↓
Segments: warm-intro-executive, warm-intro-peer, cold-with-context
   ↓
Playbook: playbooks/mutual-intro-playbook.md
   ↓
Schema: MutualIntroContext interface
   ↓
Test: Create trigger, generate sequence, approve in Slack

📚 Current Campaign Types

Campaign TypePlaybookStatusUse Case
event-follow-up✓CompletePost-conference/event outreach
mutual-intro✓CompleteWarm introduction via investor/customer
cold-outbound✓CompleteCold prospecting to ICP
partnership-outreach[ ]TODOPartner rep engagement
product-launch[ ]TODONew feature announcement
content-syndication[ ]TODOBlog/case study sharing
re-engagement[ ]TODOReactivate inactive prospects
referral-request[ ]TODOAsk for customer referrals

Your Excel tabs → Map to these or add new ones

🎓 Key Concepts

Campaign Types

Different outreach scenarios with unique triggers and goals.

Examples: event-follow-up, mutual-intro, cold-outbound

Segments

Variations within a campaign type based on context.

Example: Event follow-up segments:

  • Booth visitor (light engagement)
  • VIP dinner (high engagement)
  • Meeting log (scheduled conversation)

Personas

Prospect types with different messaging needs.

Standard 4:

  • Executive: ROI-focused, concise
  • Practitioner: Technical, tactical
  • Champion: Enablement-focused
  • Blocker: Risk-aware, proof-driven

Multi-Step Sequences

2-6 touchpoints over days/weeks with specific:

  • Timing: When to send (immediate, day-2, day-7)
  • Channel: Email, LinkedIn, phone, video
  • Goal: What this step should achieve
  • Personalization: How to customize

Human-in-the-Loop

Slack approval before sending sequences:

  • ✅ Approve - Send as-is
  • ✏️ Edit - Modify in Slack modal
  • 🔍 View Full - See complete sequence
  • ❌ Reject - Don’t send

🔧 Implementation Guide

Phase 1: Setup (Week 1)

Goal: Get basic system running with 1 campaign

  1. Setup infrastructure:

    • Next.js app with Workflow DevKit
    • Vercel AI SDK configuration
    • Slack app creation
    • HubSpot integration

  2. Implement core services:

    • Research agent (AI SDK Agent)
    • Campaign classifier (generateObject)
    • Sequence generator (generateObject)
    • Slack approval handler

  3. Add first campaign:

    • Choose simplest campaign from Excel
    • Create playbook markdown
    • Update schemas
    • Test end-to-end

Deliverable: Working system with 1 campaign type

Phase 2: Add Campaigns (Weeks 2-3)

Goal: Add 3-5 most important campaigns from Excel

  1. Prioritize campaigns:

    • Highest volume campaigns first
    • Warmest leads (highest conversion)
    • Most manual work currently

  2. Create playbooks:

    • Use Excel tabs as source
    • Follow template structure
    • Include real examples
    • Test each campaign

  3. Iterate based on feedback:

    • Review approval rates in Slack
    • Measure response rates
    • Collect edits/rejections
    • Refine prompts and templates

Deliverable: 3-5 campaign types fully automated

Phase 3: Scale & Optimize (Week 4+)

Goal: Refine system based on real usage

  1. Monitor metrics:

    • Approval rates by campaign
    • Response rates by segment
    • Time to first response
    • Meeting conversion

  2. Optimize prompts:

    • Improve research agent tools
    • Refine classification accuracy
    • Enhance sequence quality
    • Reduce edit rate

  3. Add advanced features:

    • A/B testing different sequences
    • Dynamic timing based on engagement
    • Multi-channel sequences
    • Automated follow-ups based on responses

Deliverable: Production-ready system with metrics dashboard

📊 Success Metrics

Generation Quality

MetricTargetCurrent
Approval Rate80%+-
Edit Rate<30%-
Rejection Rate<10%-
Time to Generate<2 min-

Sequence Performance

Campaign TypeOpen RateResponse RateMeeting Rate
Event Follow-Up50-60%30-40%18-25%
Mutual Intro70-80%50-60%35-45%
Cold Outbound40-50%15-20%8-12%

Efficiency

MetricTargetCurrent
Sequences/Week50-100-
Human Review Time<3 min each-
Response Time<24 hrs-

🎯 Best Practices

Playbook Writing

✅ Do:

  • Include 5+ real examples per section
  • Write actual templates (not just principles)
  • Be specific about personalization requirements
  • Include anti-patterns (what NOT to do)
  • Use real language from successful outreach

❌ Don’t:

  • Write generic templates
  • Skip personalization checklist
  • Forget to include timing guidance
  • Use vague success metrics

Personalization

✅ Do:

  • Lead with specific research insight
  • Reference recent company news/activity
  • Mention similar companies you’ve helped
  • Customize for persona (exec vs practitioner)
  • Include mutual connections when available

❌ Don’t:

  • Use generic “I loved your post” personalization
  • Copy-paste templates without customization
  • Ignore prospect’s actual challenges
  • Over-automate at cost of quality

Slack Approval

✅ Do:

  • Show full context in approval message
  • Make editing easy (modal with all fields)
  • Track why sequences are rejected
  • Send reminders after 4 hours
  • Auto-reject after 24 hours (with alert)

❌ Don’t:

  • Send approval requests to wrong channel
  • Make editing require going outside Slack
  • Ignore rejection patterns
  • Let requests sit indefinitely
  • Approve automatically without review

🛠️ Troubleshooting

Low Approval Rates

Symptoms: <70% approval rate

Causes:

  • Generic sequences (not personalized enough)
  • Poor research data
  • Wrong campaign classification
  • Templates don’t match brand voice

Fixes:

  1. Review rejected sequences for patterns
  2. Improve research agent tools
  3. Add more examples to playbooks
  4. Refine classification prompts
  5. A/B test different approaches

Low Response Rates

Symptoms: Below target response rates

Causes:

  • Wrong timing (too slow or too fast)
  • Poor personalization execution
  • Weak CTAs
  • Generic templates
  • Wrong segment classification

Fixes:

  1. A/B test timing variations
  2. Improve personalization depth
  3. Strengthen CTAs with specificity
  4. Review successful sequences for patterns
  5. Re-segment prospects

High Edit Rates

Symptoms: >40% sequences require significant edits

Causes:

  • Playbook templates too generic
  • Research agent missing key info
  • Generator prompt not specific enough
  • Brand voice not captured well

Fixes:

  1. Add more specific examples to playbooks
  2. Improve research agent with better tools
  3. Refine generator prompt with real examples
  4. Review common edits and incorporate patterns

🔗 Related Documentation

Internal Docs

External References

📞 Getting Help

Questions?

  1. Check documentation first:

  2. Review existing playbooks:

    • Look at similar campaign types
    • Study sequence structures
    • Copy patterns that work

  3. Test with sample data:

    • Create test fixtures
    • Run through full workflow
    • Check Slack approval UX

  4. Ask team for feedback:

    • Share draft playbook
    • Review approval requests together
    • Iterate based on real usage

🎉 Next Steps

For Your Specific Use Case

Based on your Excel file (GTM_Strategy_Playbooks.xlsx):

  1. Review each Excel tab - Understand what campaigns you want
  2. Prioritize top 3 - Start with highest ROI
  3. Convert to playbooks - Use HOW_TO_ADD_CAMPAIGNS.md
  4. Test with real data - Run through full workflow
  5. Iterate based on results - Monitor metrics, improve

Immediate Actions

Remember: This is a reference architecture. Adapt to your specific needs, test with real data, and iterate based on what works for your GTM motion.

📄 License

Reference implementation for Brainforge GTM team.

Graph View

Backlinks