Real-Estate AI CRM — Dashboard & Analytics

Overview

A US real-estate CRM needed its front end to actually reflect the platform underneath: a solid multi-tenant backend was in place, but much of the dashboard read from hardcoded mock data rather than the live API. The job was to turn a good backend into a product an agent opens every morning — real numbers, a working pipeline, and an AI assistant that earns its place on the screen.

The work spanned the dashboard, the analytics views and the AI-assistant surface, wiring each to the real data and giving them states that hold up when the data is empty, loading or mid-request.

What Was Built

• Command-center dashboard — active listings, new leads, pipeline value and deals won, each reading from the live API rather than placeholders
• Lead pipeline — leads by source and stage with values, moving through new / viewing / negotiation / won
• Analytics layer — leads-over-time and pipeline trends wired to the backend's reporting endpoints
• AI assistant — context-aware prompts that surface cold leads and draft follow-ups, with a cost ceiling around the model calls
• Module map across Properties, Leads, AI Agent, Analytics and Team — each audited as fully / partially / not wired, so the remaining work was a clear list rather than a guess
• Empty, loading and error states throughout, checked at desktop and phone width

Tech Stack

Outcome

The dashboard, pipeline and analytics read from real data instead of mock content, and the AI assistant went from a decorative panel to something that points at a specific lead and offers the next action. A module-by-module audit mapped exactly what was wired and what remained, turning "the front end feels unfinished" into a short, ordered backlog.

Client details and branding are withheld; the screens shown here use representative sample data.