Client Status Report
Grow Big Ventures AI Email SDR System
Current delivery status for the inbound email SDR/BDR system: what is already demo-ready, what remains for full production launch, and what access is needed from Grow Big Ventures.
Phase 1
MVP status: demo-ready
5
sample inbound replies covered
4
approval-ready draft cases
1 VPS
enough for client-owned deployment
Executive Summary
Where the system stands
The Phase 1 workflow is implemented and ready to demonstrate: inbound email reply
handling, AI classification, response drafting, lead scoring, Slack approval, and
dashboard visibility.
What is needed for full delivery
Full production launch requires deployment to GBV-owned infrastructure and live
validation against GBV-owned Instantly, GoHighLevel, Slack, and Anthropic credentials.
SOW Coverage
| SOW Capability | Status | Notes |
|---|---|---|
| Instantly inbound reply detection | Built, needs live validation | Webhook path and local readiness checks exist. Needs one controlled live Instantly campaign smoke test. |
| AI classification and response draft | Demo-ready | Five sample replies are classified with recommended actions and operator-facing reasoning. |
| Claude Sonnet / Anthropic integration | Implemented | Production should use GBV-owned Anthropic billing credentials. |
| Slack co-pilot approval | Built, needs production callback test | Slack messages can carry approve/reject actions. Editing should happen from the dashboard. |
| Conversation dashboard | Built | Conversation, approval, demo, analytics, settings, decisions, training, and follow-up views exist. |
| GoHighLevel CRM population | Built, needs GBV token test | Contact lookup/upsert, notes, tags, opportunities, tasks, and sync history are implemented. |
| GoHighLevel round-robin calendar booking | Built, needs real calendar test | Requires GBV location, calendar, user, pipeline, and stage IDs to validate in production. |
| Lead scoring and intent detection | Built | Lead score and intent facts are stored and visible in the dashboard. |
| Follow-up engine | Foundation built | Follow-up records and controls exist. Production sequences and policy tuning need GBV approval. |
| Training / sandbox mode | Demo-ready | Sandbox test cases and GBV sales guidance are loaded for controlled evaluation. |
| Autonomous mode toggle | Available with policy controls | Recommended path: start in co-pilot mode, then enable limited autonomy after review data confirms quality. |
| LinkedIn / HeyReach | Out of scope | The SOW is email-only. LinkedIn should be handled as a separate implementation track. |
Demo-Ready Today
Workflow shown in the demo
- Five inbound email replies are processed through the sandbox.
- The AI classifies intent and recommended action for each reply.
- Sendable replies get draft responses and approval status.
- Unsubscribe or disqualified replies are suppressed instead of drafted.
- Slack co-pilot approval flow is represented and ready for live callback testing.
Operator views available
- Demo replay view for the five sample replies.
- Approval queue for approve, edit, reject, and booking actions.
- Conversation inbox and lead profile visibility.
- Performance metrics including response, approval, booking, failure, and AI outcome data.
- Training and business knowledge settings for client-specific guidance.
Dashboard Navigation Map
Each dashboard area maps directly to a part of the SOW so GBV can evaluate the system as an operating workflow, not just a one-screen demo.
| Sidebar Item | What It Does | SOW Connection |
|---|---|---|
| Approvals | Shows AI-drafted replies waiting for a human to approve, edit, reject, or book. | Co-pilot mode and manual override before AI messages are sent. |
| Inbox | Displays the full conversation thread, lead context, AI replies, and operator takeover controls. | Conversations View, manual takeover, and email reply management. |
| Leads | Shows lead records, contact details, lifecycle state, score, notes, meetings, and sync history. | Lead intelligence, GHL CRM population, scoring, and centralized tracking. |
| Tasks | Lists human follow-up tasks and handoff items created when the AI needs operator help. | One human overseer model, edge-case handling, and manual intervention workflow. |
| Follow-Ups | Shows scheduled, approval-gated, blocked, cancelled, and sent follow-up actions. | Intelligent Follow-Up Engine and re-engagement when prospects do not book. |
| Demo | Runs through the five sample inbound replies and shows classification, draft, score, approval, and Slack status. | Phase 1 MVP proof: sandbox reply handling and co-pilot review. |
| Training | Shows source coverage, readiness checks, and sandbox tests for GBV-specific sales guidance. | AI Agent Customization and Training Mode. |
| Decisions | Audits AI classifications, proposed actions, policy decisions, approval state, and raw decision context. | Reliability, transparency, manual override, and safe autonomous-mode rollout. |
| Analytics | Tracks response time, booking rate, approval rate, send failures, intent mix, and AI outcomes. | Performance Dashboard and AI vs. human performance review. |
| Settings | Controls brand profile, business knowledge, integrations, policy rules, safe mode, and autonomy presets. | AI Agent Configuration, 120+ parameter direction, integrations, and co-pilot/autonomous controls. |
Needed From Grow Big Ventures
| Area | Required From GBV | Why It Is Needed |
|---|---|---|
| Client-owned hosting | One Ubuntu VPS, domain/subdomains, temporary deployment access. | Runs the API, worker, dashboard, Postgres, Redis, and HTTPS proxy under GBV control. |
| Instantly | API key, webhook secret, campaign IDs, sending mailbox IDs, and one owned test lead. | Validates real inbound reply detection and approved outbound reply sending. |
| GoHighLevel | Private Integration Token, location ID, round-robin calendar ID, user IDs, pipeline ID, and stage IDs. | Enables CRM population, lead lifecycle sync, availability checks, and appointment creation. |
| Slack | Dedicated approval channel, Slack app install permission, bot token, signing secret, and channel ID. | Enables co-pilot notifications and approve/reject button callbacks. |
| Anthropic | GBV-owned Anthropic API key. | Keeps AI usage and billing under GBV ownership. |
| Business rules | Final offer, pricing rules, objection handling, qualification rules, stop rules, and claims the AI must avoid. | Locks the AI behavior before live prospect traffic is enabled. |
| Operators | Admin and operator emails for dashboard access. | Creates the user accounts that approve drafts and handle edge cases. |
Recommended Client-Owned Deployment
Minimum infrastructure
- One Ubuntu VPS is enough for the first production launch.
- Recommended size: 4 vCPU, 8 GB RAM, 80 GB SSD.
- Docker Compose can run the full system without requiring Coolify.
- Postgres and Redis can run on the VPS; managed Postgres is optional for easier backups.
Services on the VPS
- Dashboard: operator UI.
- API: webhooks, dashboard API, approval callbacks.
- Worker: AI processing, queues, retries, follow-ups, provider sync.
- Postgres: durable source of truth.
- Redis: queues, delayed jobs, retries, and debounce timers.
- Caddy or Nginx: HTTPS and domain routing.
Security And Access Model
Dashboard authentication
The first production launch can use built-in dashboard authentication with admin and operator roles. Passwords are hashed, sessions use secure httpOnly cookies, and access is scoped by brand membership.
Provider security
Instantly webhooks use a shared secret, Slack buttons use Slack signature verification, and GHL/Anthropic credentials live only in server environment variables. No provider credentials should be stored in the public dashboard or committed to source control.
Production Launch Plan
Provision client-owned infrastructure
Prepare GBV VPS, DNS, HTTPS, Docker, Postgres, Redis, and initial dashboard users.
Connect provider credentials
Configure Instantly, GHL, Slack, and Anthropic credentials as server-side secrets.
Run controlled live smoke tests
Use owned test contacts only: one Instantly reply, one Slack approval, one GHL CRM sync, and one test calendar booking.
Begin co-pilot production
Route live replies through human approval first. Measure response time, approval rate, booking rate, and failure rate.
Expand autonomy after proof
Enable selective autonomous sending only for low-risk, high-confidence reply types after GBV confirms quality.