System Architecture — Answer BoxCONF 0.97
Direct Answer
What is the complete IO system for commercial real estate?
18 agents across 6 categories (Targeting, Enrichment, Market, Outreach, LinkedIn, Content), 13 Supabase tables, a 7-stage continuous pipeline, and a Universal CRM managing 78 properties across 3 markets. Weekly output: ~2,400 prospect scores, ~312 content packages, ~1,560 outreach touches, ~45 broker alerts. Token consumption: ~1.7M tokens/week.

Full System Scale

What does it actually look like when you put 78 commercial real estate properties under intelligent operations? Not the pitch deck version. The real version. Eighteen agents running continuously across six categories, reading from and writing to 13 database tables, scoring prospects through a 14-signal model, generating content at scale, and orchestrating outreach through a 7-stage loop that never stops cycling.

The previous nine articles in this series covered each piece in isolation. The Targeting Agent. The Enrichment pipeline. The Market Intelligence system. Prospect Qualification. Content generation. Pipeline Orchestration. Each piece is powerful on its own, but the real value emerges when they work together as a single system. The whole is not just greater than the sum of its parts — it is categorically different.

A brokerage with 78 properties and no intelligence system has 78 separate pipelines, each managed manually, each at a different stage of attention. Some properties get active outreach. Some sit dormant for weeks. The IO system eliminates this inequality. Every property has an active pipeline. Every prospect is scored. Every outreach sequence runs on schedule. The system does not forget, does not get distracted, and does not play favorites.

The question is not whether AI can manage one property’s pipeline. It is whether it can manage seventy-eight simultaneously, without any of them falling through the cracks. That is the real test.
IO System Architecture

18-Agent Inventory

The 18 agents are organized into six categories. Each category has a defined responsibility boundary, input/output contracts, and coordination protocol with other categories.

System Architecture — Agent InventoryCONF 0.96
18-Agent Inventory
6 Categories · All Active
CategoryCountAgentsPrimary Function
Targeting4ICP Matcher, Geo Expander, Market Monitor, Seed GeneratorProspect identification and qualification criteria
Enrichment12Firmographic, Behavioral, Transaction, Relationship, + 8 data normalizersData acquisition, normalization, and signal generation
Market4Vacancy Tracker, Comp Analyzer, Absorption Monitor, Demand SignalSubmarket analysis and competitive intelligence
Outreach6Email Sequencer, Phone Script, Follow-up, Meeting Scheduler, Response Handler, Cadence ManagerMulti-channel outreach execution and tracking
LinkedIn5Connection Request, Message Sequence, Profile Optimizer, Content Engagement, Network AnalyzerLinkedIn-specific outreach and relationship building
Content3Brochure Generator, Article Writer, Ad Copy EngineMarketing collateral production at scale

13 Supabase Tables

The data layer is 13 core Supabase tables, each with defined row-level security policies and real-time subscriptions. The tables are designed for high-write throughput because the pipeline generates thousands of engagement events per day across 78 properties.

The core tables are: properties (78 records with full detail), prospects (scored lead database), enrichment_data (firmographic and behavioral signals), scores (14-signal composite scores with history), segments (dynamic audience groups), outreach_sequences (multi-channel cadences), engagement_events (real-time signal tracking), content_packages (generated collateral), pipeline_runs (orchestration logs), market_data (vacancy, comps, absorption), relationship_graph (connection mapping), disqualification_flags (7-type flag system), and agent_logs (execution history for all 18 agents).

15 Property Detail Tabs

Each property in the dashboard has 15 detail tabs: Overview, Location, Financials, Tenants, Prospects, Outreach, Content, Market, Competitors, Timeline, Documents, Analytics, Notes, Scoring, and Activity. The tabs provide a complete operational view of every property without requiring the broker to switch between tools.

Universal CRM

The Universal CRM is the architectural innovation that makes the system more than a collection of property-level pipelines. Traditional CRMs are property-centric — each listing has its own prospect list. The Universal CRM tracks prospects across all 78 properties simultaneously.

A prospect interested in industrial space in DFW who also viewed office listings in Austin appears in both property pipelines with their full engagement history. This prevents the common CRE problem of the same prospect receiving conflicting outreach from different brokers at the same firm about different properties. The Universal CRM ensures one prospect, one relationship, one coordinated approach regardless of how many properties they are evaluating.

Weekly Output Metrics

Market Agent — Weekly MetricsCONF 0.95
Weekly System Output
~2,400
Prospect Scores
326 prospects x multiple cycles
~312
Content Packages
4 outputs x 78 properties
~1,560
Outreach Touches
Email + LinkedIn + Phone
~45
Broker Alerts
On Fire + Hot promotions
~12
Segment Rebuilds
Dynamic audience updates
1.7M
Tokens / Week
All agents combined

The Future

The system described in this series is not the endpoint. It is the foundation. The next phase adds predictive deal scoring — using historical transaction data to predict not just which prospects will engage, but which will close and at what price point. The phase after that adds autonomous market response — the system detecting a shift in submarket dynamics and automatically adjusting property positioning, content strategy, and outreach messaging without human instruction.

But those are future articles. What exists today — 18 agents, 13 tables, 78 properties, a continuous 7-stage pipeline, and a Universal CRM — is already transforming how commercial real estate operates. Not by replacing brokers, but by giving them an operating system that makes every property visible, every prospect scored, and every outreach cycle informed by the last one.

The goal was never to build an AI that does real estate. The goal was to build an operating system that makes real estate professionals better at every part of their job, simultaneously, across every property they manage.
IO Platform — The Complete System
System — Lead CaptureCONF 0.94
IO Platform · Complete System
Get the complete system architecture document.
18-agent inventory, 13-table schema, 7-stage pipeline configuration, Universal CRM specification, and weekly output benchmarks.
Free. No spam. Unsubscribe anytime.
System — FAQsCONF 0.96

Frequently Asked Questions

5 Questions
How many agents run in the complete IO system?
18 agents in 6 categories: Targeting (4), Enrichment (12), Market (4), Outreach (6), LinkedIn (5), and Content (3). Each category has defined responsibility boundaries and coordination protocols.
Structured as FAQ schema (JSON-LD) for AEO indexing
What Supabase tables power the system?
13 core tables: properties, prospects, enrichment_data, scores, segments, outreach_sequences, engagement_events, content_packages, pipeline_runs, market_data, relationship_graph, disqualification_flags, and agent_logs. All with RLS policies and real-time subscriptions.
What are the 15 property detail tabs?
Overview, Location, Financials, Tenants, Prospects, Outreach, Content, Market, Competitors, Timeline, Documents, Analytics, Notes, Scoring, and Activity. Complete operational view per property without tool-switching.
What does the Universal CRM do?
Tracks prospects across all 78 properties simultaneously. A prospect interested in industrial in DFW and office in Austin appears in both pipelines with full history. Prevents conflicting outreach from different brokers about different properties.
What are the weekly output metrics?
~2,400 prospect scores, ~312 content packages, ~1,560 outreach touches, ~45 broker alerts, ~12 segment rebuilds. Total token consumption: ~1.7M tokens/week across all 18 agents.