Instagram AI Consultant: −3 Hours/Day of Routine

The Store: 3 People Drowning in DMs

This online store sells fashion accessories through Instagram — their primary sales channel. With 15,000+ followers and growing, they were getting 40-60 DM inquiries per day. Three team members handled everything: DM responses, order processing, inventory management, packaging, and shipping.

The math was brutal. Each DM conversation took an average of 5-7 minutes (multiple back-and-forth messages). At 50 inquiries per day, that's over 4 hours consumed by DMs alone. And 80% of those conversations were the same questions repeated endlessly.

The Message Audit: What Customers Actually Ask

We analyzed 500 DM conversations over two weeks and categorized every inquiry:

90% of inquiries could be answered with information already available — product catalog data, shipping tables, and store policies. The team was essentially acting as a human FAQ page for 3+ hours every day.

Worse, response times during business hours averaged 4-6 hours (because the team was busy with other tasks), and after-hours messages got zero response until the next morning. In e-commerce, a delayed response often means a lost sale.

Building the AI Consultant

We designed an AI consultant with three distinct conversation funnels, each optimized for a different customer type:

Funnel 1: New Customer. When a new follower sends their first DM, the AI greets them, asks what they're looking for, and provides personalized product recommendations based on their stated preferences. It can share product photos, prices, available sizes, and direct links to order. If the customer is interested, it guides them through the purchase process.

Funnel 2: Returning Customer. Recognized customers (matched by Instagram username against the order database in Google Sheets) get a different experience. The AI knows their purchase history and can suggest complementary items. It handles re-orders, size exchanges, and loyalty-based offers.

Funnel 3: Post-Purchase Support. Customers who have active orders can ask about delivery status, tracking numbers, return procedures, and exchange policies. The AI pulls real-time data from the order management sheet.

The Technical Setup

The system runs on three components:

• Claude API — powers the conversational AI with product knowledge, store policies, and brand voice
• n8n — orchestrates the workflow: receives Instagram webhook events, routes to appropriate funnel, calls Claude API, sends responses back through Instagram Graph API
• Google Sheets — serves as the product catalog and order database (the store was already using it), keeping the AI updated with real-time inventory and pricing

We trained the AI on the full product catalog (200+ items), store policies, shipping tables for all regions, and 50 example conversations that showed the ideal tone and response style. A key design decision: the AI never pretends to be human. It introduces itself as an AI assistant and transparently escalates to the human team when it encounters questions it can't handle confidently.

The Escalation Protocol

Not everything should be automated. We built clear escalation rules:

• Custom orders or bulk inquiries → forwarded to the team with full conversation context
• Complaints or negative sentiment → immediate team notification via Telegram
• Customer explicitly asks to speak to a human → instant handoff with conversation summary
• AI confidence below threshold → "Let me check with the team" response, forwarded for human review

Results: Before vs. After