Client Success
01
How It Works
This shows how EyeFly monitors all active clients for risk signals — from raw Slack messages and GHL bot scores through automated pipelines into Postgres, surfaced to the CSM team via a daily inbox and live kanban dashboard.
02
Layer 1 — Native Apps
LAYER 1 — NATIVE APPS
Where client data and issues surface
Slack
- Each client has a private support channel
- Team flags issues, complaints & bot errors here
- Primary async surface for client comms
GoHighLevel
- Bot conversations with leads happen here
- Lead qualification, disqualification & transfer
- Appointment outcomes tracked per client
Google Sheets
- Leads Transferred sheet updated after every transfer
- CSM tracks notes & follow-ups here
- Manual override log for edge cases
Meta Ads
- CPL and spend trends feed health scoring
- Underperforming ad accounts trigger risk flags
- Attribution verified via CAPI events
03
Layer 2 — Connectors
LAYER 2 — CONNECTORS
Automated pipelines — always running
Scheduled Tasks
- Client health scored nightly at 11 PM
- Client risk kanban rebuilt daily at 11 AM
- Slack DMs swept for action items 3× daily
Slack MCP
- Reads all client support channels
- Sends P0/P1 alerts instantly to Akash
- Posts CSM daily digest to #akash-notes
GHL MCP
- Pulls bot conversation scores for QA
- Reads disqualification error patterns
- Source for daily bot error classification
Health scoring and ticket capture run unattended. CSM inbox assembled every morning.
04
Layer 3 — Data Storage
LAYER 3 — DATA STORAGE
Everything persisted in Postgres
Support Tickets
- Every Slack complaint logged with severity P0–P3
- Tracks resolution status and owner
- Escalation counter per recurring issue type
Client Health
- Rolling score: CPL trend + bot QA + tickets
- Flags clients at churn risk before they cancel
- Updated every night, read by CSM inbox daily
Bot QA Scores
- Daily pass/fail score per client bot
- Logs ideal-flow deviations with conversation ID
- Feeds health score and CSM daily checklist
Complaint Log
- Raw client feedback classified into 10 types
- Auto-routed to right owner as a task
- Escalation tracker detects recurring patterns
05
Layer 4 — Skills
LAYER 4 — SKILLS
Claude intelligence layer — on-demand or scheduled
Support Tickets Review
- Scans all client Slack channels for new issues
- Classifies severity P0–P3 with reasoning
- Creates a routed task for each ticket
Client Health Check
- Surfaces at-risk clients with root cause
- Recommends specific retention action per client
- Runs on-demand or inside CSM inbox
CSM Inbox
- Regina's daily to-do: tickets + health + bot QA
- Single list instead of checking four tools
- Generated each morning, reviewed in Claude Chat
Escalation Tracker
- Detects when same issue type recurs ≥2×
- Promotes repeated patterns into SOP projects
- Prevents the same fire from being fought twice
06
Layer 5 — Frontend
LAYER 5 — FRONTEND
What the team sees and uses
Client Kanban
- Live risk board: green / yellow / red per client
- Rebuilt every morning from overnight health scores
- Shareable link for the full CSM team
Slack Alerts
- P0/P1 tickets fire instantly to Akash DM
- Daily CSM digest posted to #akash-notes
- Escalation pattern alerts at threshold
Claude Chat
- Ask 'who is at risk this week?' for instant report
- Run CSM inbox on demand
- Investigate any client's full history
07
Vision
VISION — COMING NEXT
Proactive Churn Prevention
Client health score drops → CSM auto-alerted with a recommended action before the client ever complains. At 330 clients, 5% monthly churn = $27K MRR lost.