Automated Sales Workflow – Real Estate Case Study

Executive Summary

• Client context: Real estate sales team handling inbound Instagram enquiries
• Sector: Property sales, AI automation, lead qualification
• Engagement: Workflow design, prompt logic, database modelling, integrations, moderation flow
• Core stack: n8n, PostgreSQL, OpenAI, vector database, Instagram API, Gmail, Google Drive
• Outcome snapshot: A multi-stage sales workflow that captures inbound leads, replies instantly, schedules follow-ups automatically, and keeps humans in control of approvals when needed

The Problem

In real estate, slow response times destroy pipeline value. Agencies often depend on agents manually checking Instagram messages, qualifying interest, sending the first reply, remembering follow-ups, and pulling the latest listing information from multiple places.

That creates predictable failure points:

• hot leads wait until someone is available
• follow-ups happen inconsistently
• staff repeat the same qualifying work every day
• sales messages drift away from approved wording and current listing data
• management has weak visibility into what stage each enquiry has reached

The client needed a system that could respond at any time, stay aligned with current property information, and still allow human review when appropriate.

Why It Was Difficult

Automating sales is not just a chat problem.

The solution had to combine:

• live inbound messaging from Instagram
• context-aware response generation instead of canned replies
• stage-based workflow logic for follow-up timing
• structured storage of chat history and lead state
• current listing knowledge that could be refreshed without redeploying the whole system
• optional human moderation so the business could keep commercial control

If any of those layers failed, the workflow would feel either brittle, unsafe, or obviously robotic.

Our Solution

We designed a modular automation system around n8n as the orchestration layer.

Incoming Instagram messages are received via webhook, stored in PostgreSQL, and enriched with conversation history through a memory layer. Each message is then processed through OpenAI-powered response generation, using prompt logic plus historical context to create a reply that fits the current stage of the conversation.

Rather than allowing AI to operate as an ungoverned black box, we built an approval path through Gmail notifications so managers or developers can approve, edit, or reject replies before they are sent in workflows that require oversight.

We also added a vectorized knowledge layer and document update flow via Google Drive, allowing the system to stay aligned with changing listings and policies without rewriting the workflow from scratch.

What We Delivered

• Instagram DM capture via webhook
• PostgreSQL lead and chat-history storage
• GPT-based response generation with prompt context and memory
• JSON-formatted outbound delivery flow for stable message handling
• Gmail-based moderation and approval path
• Stage-based follow-up automation using scheduled triggers
• Vector search for listing and policy knowledge retrieval
• Google Drive content refresh path for operational updates

Results

• Always-on first response: every inbound Instagram enquiry became eligible for immediate system handling instead of waiting for agent availability
• Cleaner sales discipline: follow-ups moved from ad hoc reminders to stage-based automation tied to the lead lifecycle
• Lower manual workload: staff no longer need to watch every DM thread from first contact; their effort shifts toward exception handling and deal progress
• Better control than a fully automatic chatbot: Gmail moderation preserved human override for sensitive or higher-value conversations
• Reusable architecture: the workflow can be extended to other platforms and sales environments without redesigning the entire logic stack

Technical Highlights

Workflow orchestration in n8n

n8n allowed us to express the commercial process as an actual system rather than a loose set of scripts. That matters because sales automation usually fails at the handoffs between triggers, AI, data, and approvals.

Memory plus structured storage

PostgreSQL and chat memory prevent the common chatbot problem of answering each message as if it were the first one. The system keeps state, which is essential for stage-based sales conversations.

Dynamic knowledge updates

The knowledge layer is not static. By updating source material from Google Drive and vectorizing it for retrieval, the team can keep the AI aligned with current listings and policies without rebuilding the application.

Proof Artifacts

• Portfolio asset: /public/assets/portfolio/realestate-sales.webp
• Internal proof points captured in the system design include Instagram lead capture, Gmail moderation, vector search, and Google Drive knowledge refresh
• Related live capability on the site: AI sales and intake examples

Transferable Value For Clients

This case study is relevant if you need:

• AI lead qualification for social or inbound channels
• faster first response without hiring a round-the-clock team
• human-in-the-loop review for commercial or compliance reasons
• workflow automation tied to sales stages rather than one-off replies
• a CRM-like operational layer without forcing agents into extra manual admin

The transferable lesson is simple: AI becomes commercially useful when it is embedded in a controlled workflow with memory, data, scheduling, and oversight.

If you want a related example focused on turning enquiries into structured offers and project estimates, see Vasilkoff.info.

Work With Us

If you need an AI workflow that responds faster, follows up consistently, and still stays under business control, contact us. We build automation around measurable sales outcomes, not demo-chat novelty.