Eliminating lead leakage by piping inbound messages from Instagram DMs, WhatsApp, and email into a single AI dispatcher that instantly answers FAQs, qualifies prospects, and autonomously books meetings directly into your sales team's calendar.
Zero
lead leakage
across all channels
< 8s
response time
FAQ auto-replies
24/7
coverage
no human needed
3–4 wk
build sprint
fixed price
Typical build: 3–4 week sprint · Fixed price · Zero delivery risk
Channels
Instagram · WhatsApp · Email
FAQ reply
Under 8 seconds
Booking
Leads scoring ≥7
Instagram DMs land with the social team. WhatsApp goes to someone's personal phone. Email hits a shared inbox no one owns. A lead who messages on all three channels gets three different experiences — or none at all.
Studies consistently show that responding to a lead within 5 minutes is 100x more effective than responding after 30 minutes. A manually monitored inbox makes sub-5-minute response impossible at any scale.
Pricing questions, service scope questions, turnaround time questions — the same 10 questions asked 50 times a week. Every one of those responses is a sales rep's time that could go to qualified opportunities.
This is the actual workflow Kovil AI builds and deploys — not a diagram. Here is what runs inside every node.
Messages arrive via Instagram DM, WhatsApp Business, and email simultaneously. n8n webhook listeners capture all three channels in real time, normalising each message into a unified payload with: channel source, sender ID, message text, timestamp, and any attachments. Every inbound message — regardless of origin — enters the same pipeline from this point forward, eliminating siloed inboxes and ensuring zero message is missed.
GPT-4o classifies the message into one of four buckets: FAQ (question about pricing, services, turnaround), New Lead (expressed buying intent), Existing Client (reference to current work), or Spam/Irrelevant. Classification is done via a structured JSON output prompt — no free-text parsing required. This means downstream routing is deterministic: a JSON key, not a regex pattern, decides what happens next. Misclassification rate is under 2% in production.
For FAQ classifications, GPT-4o pulls the relevant answer from a Notion knowledge base page (cached in n8n memory every 6 hours) and replies instantly in the same channel the message came from — Instagram DM reply, WhatsApp message, or Gmail reply. Average response time: under 8 seconds. Answers are contextualised to the specific question phrasing — not templated responses. The Notion cache prevents GPT-4o from hallucinating answers on topics not in your KB.
For new leads, GPT-4o sends three qualifying questions tailored to the detected channel: budget range, project type, and timeline. Responses are collected asynchronously — n8n waits for all three before continuing. Once all three are answered, a lead score (1–10) is calculated based on predefined criteria. This flow handles multi-turn conversations natively — if the lead responds across 3 separate messages, n8n stitches them into a complete qualification record.
Leads scoring ≥7 receive a personalised Calendly booking link with a contextual message ("Based on what you've shared, a 30-minute discovery call would be the best next step — book a time here"). The Calendly link uses UTM parameters to track which channel the booking came from, giving your team full attribution data. Leads scoring below 7 receive a nurture message appropriate to their channel and are flagged in HubSpot for manual follow-up.
Every interaction is logged: contact created or updated in HubSpot with all conversation data, lead score, channel source, qualification answers, and booking status. Existing contacts are matched by email or phone and updated rather than duplicated. The workflow creates a custom HubSpot timeline event for each interaction, giving your sales team a complete chronological view of every touch — from first Instagram DM through to booked meeting — in one CRM record.
DM capture
Webhook listens on the Instagram Messaging API. All DMs normalised to the unified payload format before routing.
WhatsApp capture
WhatsApp Business webhook via Meta Cloud API. Supports text, images and documents from the start.
Email capture
Gmail push notification (Pub/Sub) triggers n8n on new emails to your business inbox. Replies sent from the same thread.
Intent classifier + qualifier
Two roles: (1) classify intent via structured JSON; (2) generate contextual FAQ answers and qualifying questions per channel.
Orchestration
Manages all webhooks, routing logic, async multi-turn conversation state, memory caching, and parallel output branches.
Meeting booking
Personalised booking links with UTM parameters. Routing rules ensure leads land on the correct calendar and meeting type.
CRM sync + attribution
Contact deduplication by email/phone, custom timeline events per interaction, lead score field, and booking status tracking.
Kovil AI scopes, builds, tests and deploys this workflow end-to-end. You do not touch n8n until it is live across all three channels.
GPT-4o's intent classification includes a confidence score. If the message is ambiguous or classified as Existing Client with a complex query, the workflow routes it to a human handoff queue in Slack — tagged with the original message, channel source, and contact history. No message is ever dropped.
Yes. GPT-4o detects the language of each incoming message and responds in the same language. The Notion knowledge base can contain multilingual FAQ entries, and GPT-4o will match the response language to the sender automatically.
Leads scoring below 7 receive a nurture message — typically a case study or resource link — and are logged to HubSpot with their score and answers. A follow-up reminder is created in HubSpot for the account manager to review and manually re-engage.
Both. The workflow checks HubSpot for an existing contact matching the sender's email or phone before creating a new record. If a match is found, the interaction is logged as a new timeline event on the existing contact — enriching the record rather than duplicating it.
Book a 30-minute discovery call. We will scope the dispatcher for your three channels, your knowledge base and your HubSpot setup — fixed price, zero delivery risk.
Typical sprint: 3–4 weeks · Fixed-price · Fully managed delivery · Post-launch support included