Real Estate Operating System case study

Confidential (real estate group) · Real estate

A multi-tenant AI platform automating real estate agency operations. Three specialised agents share one workflow across 30+ agencies.

Private deployment, client project

Three specialised agents, each scoped to one job.

Instead of one general assistant handed everything, CRM, compliance and knowledge work are split across three agents, with LangGraph coordinating the hand-offs. Agency differences sit behind CRM integrations, so more than 30 agencies share one workflow instead of thirty variants.

What keeps it lean

01One agent per responsibility
02Explicit hand-offs between agents
03One shared workflow for 30+ agencies

Rules and ordinary software before AI. A model only where it earns its place.

The challenge

Estate agencies ran CRM, compliance, marketing and documents across disconnected tools and manual processes, and each agency did it slightly differently. There was no shared system for clients and properties, and agreements were retyped from scratch. The group needed one platform that fitted thirty agencies without becoming thirty products.

The work

We built a multi-tenant platform with three specialised agents, for CRM, compliance and knowledge, coordinated by LangGraph, rather than one general assistant handed everything. Agency differences sit behind CRM integrations so the workflow stays shared, and a marketing module and a document generation and e-signing pipeline complete the operating system.

Inside the product

More than one screen.

3 more screens from Real Estate Operating System, captured at device size.

Leasing pipeline

Leads scored by the CRM agent and moved through viewing, offer and contract.

Agreement generation

Tenancy agreements generated from templates and sent for e-signature.

AI agents

CRM, compliance and knowledge agents scoped to each agency’s own data.

On the phone

Home

Leasing pipeline

Agreement generation

AI agents

How it is built

Architecture

A React frontend talks to a Python and FastAPI backend, while LangGraph orchestrates specialised multi-agent workflows powered by OpenAI models. The multi-tenant architecture serves more than 30 agencies from one platform while supporting each agency’s own CRM and workflows.

Key features

CRM agent syncing with agency-specific CRMs

Compliance agent for regulatory workflows

Knowledge agent for pipeline and workflow Q&A

Marketing module for photography booking and brochure ordering

Vendor reporting

Automated document generation and e-signing

Challenges & solutions

Challenge

Each agency used its own CRM and operating workflow, so one fixed integration could not serve every tenant.

Solution

Put the differences behind agency-specific CRM integrations while keeping the agent workflow and product experience shared across tenants.

Challenge

CRM, compliance, knowledge and document tasks had to cooperate without turning one long workflow into a fragile chain.

Solution

Split the work into specialised agents and used LangGraph to coordinate their responsibilities and hand-offs.

The impact

What became easier, better or more effective.

More than thirty agencies now run CRM, compliance, marketing and paperwork on one platform. Tenancy agreements generate from templates and go straight to e-signature instead of being retyped, and the agents handle the routine work that used to sit in inboxes, so people spend more of their day with clients.

30+

agencies on one platform

Each with its own CRM integration and workflows, served from a single multi-tenant deployment, so an improvement reaches everyone at once.

70%

less time on documents and agreements

Tenancy agreements generate from templates and are sent for e-signature without a person retyping them.

specialised AI agents

CRM, compliance and knowledge agents scoped to each agency’s own data and coordinated by LangGraph, each doing one job well.

Stack