Skip to content
Qzentra

Case fileAI Systems & Automation

Multimodal AI Sales Funnel

An end-to-end sales funnel for a crypto-mining hardware enterprise: Instagram and WhatsApp leads qualified, answered from a catalog of 500+ ASIC miners, scored, and handed to sales.

500+ASIC miners in the product catalog the system answers from

Enters through
01AI Systems & Automation
System type
Messaging-channel sales automation with multimodal AI
Also demonstrates
AI agents · RAG · messaging APIs · data integration · human handoff03AI Agents & Voice AI
System route: Instagram intake to handoff and follow-up.Eight stages across five layers. The route rises to the handoff and returns to automation for follow-up or archiving.06  Application05  Automation04  AI03  APIs02  Infrastructure01  Data12345678
  1. 1Instagram DMsAPIs
  2. 2Lead qualificationAutomation
  3. 3WhatsAppAPIs
  4. 4Multimodal AI (text, voice notes, images)AI
  5. 5Product knowledge retrievalData
  6. 6Lead scoringAutomation
  7. 7Human sales handoffApplication
  8. 8Automated follow-up or cold-lead archivingAutomation

01 Context

The challenge

A crypto-mining hardware enterprise received sales enquiries through Instagram direct messages. Each one had to be answered with accurate product knowledge across a catalog of more than 500 ASIC miners, qualified, and passed to a salesperson when it was worth their time.

Prospects did not write tidy text. They sent voice notes and photographs of hardware alongside messages, and the conversation moved from Instagram to WhatsApp before a sale. The requirement was to handle those conversations without a person reading every message, while keeping product answers accurate and getting real buyers to sales.

What made it non-trivial

  • Three input modes in one conversation: text, voice notes, and images
  • Product answers had to come from the real catalog, not from the model’s memory
  • The conversation crossed two messaging channels
  • Only qualified leads should reach a person, and every other lead needed a defined path

02 The system

What Qzentra engineered

Qzentra built the complete funnel as one production workflow, orchestrated in n8n. Instagram direct messages enter the system and are qualified; the conversation continues on WhatsApp, where a multimodal AI layer handles text, voice notes, and images.

Product questions are answered by retrieval over an Airtable-backed catalog of 500+ ASIC miners, so answers come from the catalog rather than from the model. Each lead is scored; qualified leads are handed to a salesperson, and the rest enter timed follow-up or are archived as cold.

What Qzentra was responsible for

  1. Messaging integration for Instagram DMs and WhatsApp
  2. Lead qualification and scoring logic
  3. Multimodal input handling: text, voice notes, images
  4. Retrieval over the Airtable product catalog
  5. Human handoff to sales, timed follow-up, and cold-lead archiving
  6. Workflow orchestration in n8n

03 The route

How the system works

Each stage of the workflow, on the layer where it runs. The conversation crosses two channels and comes back to automation after the handoff.

  1. Instagram DMs

    APIs

    Inbound direct messages are the entry point. Each new conversation is captured through the Instagram messaging integration.

  2. Lead qualification

    Automation

    Qualification logic runs before the conversation goes any further, so only real sales enquiries continue.

  3. WhatsApp

    APIs

    Qualified conversations continue on WhatsApp, the channel the sale is closed on.

  4. Multimodal AI (text, voice notes, images)

    AI

    The AI layer understands text, voice notes, and images, so a photo of a machine or a spoken question is handled in the same conversation as typed messages.

  5. Product knowledge retrieval

    Data

    Product questions are answered by retrieval over the Airtable-backed catalog of 500+ ASIC miners, not from the model’s memory.

  6. Lead scoring

    Automation

    Each lead is scored as the conversation develops.

  7. Human sales handoff

    Application

    Leads that score highly are routed to a salesperson. The system stops; a person continues.

  8. Automated follow-up or cold-lead archiving

    Automation

    Leads that are not ready enter timed automated follow-up. Leads that go cold are archived, so nothing sits unresolved in the channel.

04 Decisions

Engineering decisions

  1. One orchestrated workflow, not a chain of tools

    The whole funnel runs as a single n8n workflow, so qualification, messaging, AI, retrieval, scoring, and handoff share one state and one place to change the logic.

  2. Qualify before the channel moves

    Qualification happens on Instagram, before the conversation moves to WhatsApp, so only qualified conversations are carried into the closing channel.

  3. Retrieval over the catalog, not model memory

    Product answers are grounded in the Airtable catalog through retrieval. A catalog of 500+ miners changes; a model’s training data does not.

  4. Multimodal input as a first-class path

    Voice notes and images are processed alongside text rather than rejected or diverted to a person.

  5. A person at the point of value

    Scoring decides when a salesperson is brought in. Everything before that point is automated; the handoff itself is deliberate and human.

  6. Every lead has an exit

    Leads that are not ready receive timed follow-up; leads that go cold are archived. The channel does not accumulate unresolved conversations.

05 Verified outcomes

What was achieved

  • Inbound Instagram and WhatsApp leads are qualified without a person reading every message
  • Text, voice notes, and images are handled in one conversation
  • Product answers are drawn from the catalog of 500+ ASIC miners
  • Qualified leads reach sales through a structured handoff; the rest are followed up automatically or archived
500+ASIC miners in the product catalog the system answers from

06 Technology and capabilities

Technology, after the architecture

Named last, with the role each played. The capabilities are what the project demonstrates beyond its primary service.

  • n8nWorkflow orchestration for the whole funnel
  • Instagram messagingInbound channel
  • WhatsAppConversation and closing channel
  • AirtableProduct catalog store, 500+ ASIC miners
  • RAGProduct knowledge retrieval over the catalog
  • Multimodal AIUnderstanding text, voice notes, and images

Capabilities demonstrated

  • n8n workflow orchestration
  • messaging integration
  • multimodal input handling
  • RAG
  • lead qualification and scoring
  • sales routing
  • automated follow-up
  • human-in-the-loop workflow

07 Related work

Other case files

Systems that share a layer, a capability, or a decision with this one.

  1. 02Case fileAI Systems & Automation

    AI Website Audit & Outreach Engine

    An automated pipeline from a list of target URLs to personalized, segmented outreach: crawling, contact extraction, PageSpeed analysis, LLM-written messages, and automated follow-ups.

    Verified factEvery message grounded in a Google PageSpeed audit of the prospect’s site

    Enters through01AI Systems & Automation

    Read the case file
    AI Website Audit & Outreach Engine: route through the six layers.06  Application05  Automation04  AI03  APIs02  Infrastructure01  Data1234567
    1. 1Target URLsData
    2. 2Crawling and contact/social extractionData
    3. 3Google PageSpeed APIAPIs
    4. 4Technical analysisAI
    5. 5LLM-generated personalized outreachAI
    6. 6SegmentationAutomation
    7. 7Automated follow-upsAutomation

  2. 05Capability caseAI Agents & Voice AI

    Voice AI Systems

    Voice systems on real telephony: AI receptionists, inbound support, outbound qualification, appointment booking, and CRM-integrated calling, with transcription, call analysis, and escalation to people.

    Enters through03AI Agents & Voice AI

    Read the case file
    Voice AI Systems: route through the six layers.06  Application05  Automation04  AI03  APIs02  Infrastructure01  Data1234567
    1. 1Telephony: inbound or outbound callAPIs
    2. 2Speech recognitionAI
    3. 3LLM conversation logicAI
    4. 4Tools and CRM actionsAutomation
    5. 5Speech synthesisAI
    6. 6Transcription and call analysisData
    7. 7Escalation and human transferApplication

08 Discuss a project

Have inbound conversations that a person is still triaging?

Describe the channels, the knowledge the answers must come from, and where a salesperson should take over. That is enough to scope the system.

Discuss a ProjectAll work