GHL Ops & Onboarding
01
How It Works
This shows how EyeFly manages all GHL operations — from initial client onboarding through daily bot QA, lead tracking, and tag reconciliation. Every new client flows through this stack before their first lead arrives.
02
Layer 1 — Native Apps
LAYER 1 — NATIVE APPS
Where lead and client data lives
GoHighLevel
- CRM: every lead, contact & conversation
- Bot qualifies leads 24/7 via Conversation AI
- Pipeline tracks each lead from inquiry to transfer
Meta Ads
- Messenger ads drive leads into GHL bots
- Business Manager holds all page connections
- CAPI dataset receives conversion events
Google Sheets
- Leads Transferred sheet is the ops source of truth
- Cross-checked against GHL tags nightly
- Drift flagged as an anomaly automatically
Granola
- Meeting transcripts for client kick-offs
- Action items extracted and saved to brain
- Speaker-labeled notes available for recall
03
Layer 2 — Connectors
LAYER 2 — CONNECTORS
Automated pipelines — always running
n8n
- Syncs all GHL conversations to brain nightly
- Delta-syncs lead contact tags at 3 AM
- Reconciles 'transferred' tags vs Leads CSV at 7 AM
GHL MCP
- Creates subaccounts, snapshots & custom values
- Updates bot prompts and pipeline stages
- Reads Conversation AI agent config
Scheduled Tasks
- Daily bot QA scoring at 9 AM M–F
- Daily disqualification error scan at 9 AM
- GHL pipeline snapshot every morning at 5 AM
ETL runs unattended. GHL write-ops require explicit approval before Claude executes.
04
Layer 3 — Data Storage
LAYER 3 — DATA STORAGE
Everything persisted in Postgres
Lead Contacts
- Every contact across all client subaccounts
- Tag history: when 'qualified' and 'transferred' fired
- Delta-synced — only new/changed contacts fetched
PIT Registry
- Phone Inbox Token per client subaccount
- Required for all CAPI events and bot reads
- Stored once, never re-fetched
Client Directory
- All clients with GHL subaccount and ad account IDs
- Niche, page ID, CAPI dataset per client
- Single source of truth for all automations
Bot QA Scores
- Daily score per client bot with deviation log
- Ideal-flow violations with conversation ID
- Feeds client health and CSM inbox
05
Layer 4 — Skills
LAYER 4 — SKILLS
Claude intelligence layer — on-demand or scheduled
New Client Onboarding
- Zero-to-live: subaccount → snapshot → CVs → CAPI
- Interactive wizard with Claude owning 6 of 10 steps
- Target: client live in under 2 hours
GHL API
- Update bot prompts for any client by name
- Create/inspect pipelines and custom fields
- Full subaccount management from Claude Chat
CV Sync
- Pushes all 17 Client Info fields to GHL
- Fills niche-specific defaults automatically
- Idempotent — only updates fields that drifted
Bot Conversation Audit
- Reads and scores bot conversations for quality
- Detects double-asks, wrong DQs, payload errors
- Only skill allowed to read GHL conversation data
CAPI Test Fire
- Fires test LeadSubmitted event to Meta dataset
- Verifies attribution pipeline is connected
- Run after every new client goes live
06
Layer 5 — Frontend
LAYER 5 — FRONTEND
What the team sees and uses
Claude Chat
- Run full onboarding wizard step by step
- Update any client's bot prompt by name
- Audit bot conversations for any time window
GHL Live
- Subaccount created and live in GHL UI
- Bot running and qualifying leads in real-time
- Pipeline tracking every lead's stage
Slack Alerts
- Tag reconciliation drift fires to Akash
- Bot disqualification errors flagged daily
- Onboarding completion confirmed on finish
07
Vision
VISION — COMING NEXT
Zero-Touch Onboarding
New client signed → subaccount created, CVs pushed, snapshot applied, CAPI verified, first campaign live — fully autonomous. Target: under 2 hours, zero manual steps.