Terrell A. Lancaster
Back to work
Case Study

Agent Rex AI — Real Estate Command Center

An autonomous, multi-provider AI command center for real estate operations — lead parsing, market insight, scheduling, and CRM in a single pane.

Agentic AIReal EstateMulti-Model LLMGCPNext.js
Multi-LLM
AI models
90-98%
Accuracy
GCP
Platform

The Problem

Real estate professionals juggle disconnected tools — lead intake forms, MLS data, CRMs, schedulers, market reports — and none of them talk to each other. The work is repetitive: parse a lead, pull comps, score the opportunity, book the follow-up, log it all. Done by hand, it bleeds hours every day and the high-value work (pricing strategy, forecasting, relationship building) gets the leftovers.

The goal was to collapse that stack into one autonomous assistant: a single pane where an agent reads the lead, reasons over market data, drafts the response, updates the CRM, and schedules the next touch — without a human in every loop.

Architecture

A multi-provider AI command center. A multi-model router fans work across Gemini, Anthropic, and OpenAI for the task each does best; Google Genkit handles document understanding, and everything persists to Firestore on GCP behind a Next.js 15 interface.

Interfaces
Single-pane consoleCRM workflowsScheduling
AI pipeline
Multi-model routerGoogle Genkit document processingLead scoring
Capabilities
Market analysisProperty valuationPredictive forecasting
Runtime
Next.js 15 + TypeScriptGCP infrastructureFirestore

Approach

  • Built a multi-model AI pipeline that routes each task to the provider best suited for it — Gemini, Anthropic, and OpenAI behind a single orchestration layer — so property valuation, lead scoring, and market forecasting each run on the model that handles them best.
  • Used Google Genkit for intelligent document processing, turning unstructured intake (lead forms, listing sheets, comps) into structured records the CRM and analysis layers can act on.
  • Automated the CRM workflow end-to-end — lead parse, score, insight, follow-up scheduling — so the repetitive operational layer runs without a human in every loop.
  • Delivered it as a Next.js 15 + TypeScript application on GCP with Firestore as the system of record, keeping the whole command center in one pane.

Outcomes

A single command center where an autonomous, multi-provider AI assistant handles lead parsing, market analysis, document processing, and CRM workflows for real estate professionals. The multi-model pipeline delivers 90-98% accuracy across property valuation, lead scoring, and predictive market forecasting — collapsing a disconnected toolchain into one pane and freeing the operator to focus on strategy and relationships.

Stack
Next.js 15TypeScriptGoogle GenkitGeminiAnthropicOpenAIGCPFirestore
    Ask Terrell's AI