BRIX

AI that powers modern property operations with end-to-end automations, saving managers millions and unlocking true portfolio scale.

Comment

BRIX

Inspiration

Property managers process enormous volumes of maintenance invoices, work orders, and vendor interactions every month. Most of this work is manual, time-consuming, and nearly impossible to audit at scale.

Common issues we observed:

  • Vendors overbilling because no one has time to verify.
  • Thousands of invoices with almost no visibility.
  • Owners demanding proof of savings and justification of spend.

Billions are lost annually due to padded invoices, scope misalignment, and unverified work.
There was no system connecting invoice → work order → photos → pricing benchmarks into one intelligent workflow.

BRIX was inspired by a simple question:
What if property managers had an AI system that could audit everything, verify work, catch waste, and help portfolios scale?

What it does

BRIX is an AI platform that automates maintenance auditing for property management teams.

It can:

  • Ingest invoices, work orders, and before/after photos.
  • Parse invoices into structured line items and labor entries.
  • Benchmark pricing using vendor profiles and market heuristics.
  • Verify work completion through image analysis.
  • Cross-check scope against the original work order.
  • Flag padded, duplicate, or suspicious charges.
  • Recommend approve / dispute / adjust decisions.
  • Suggest owner vs. tenant charge classifications.
  • Identify better vendors for future work.
  • Generate portfolio-level savings dashboards and vendor scorecards.

Forward any invoice → receive analysis, evidence, and quantified savings.

How we built it

BRIX is built as an end-to-end multimodal AI system with three core layers:

1. Ingestion

  • Email forwarding and file uploads
  • OCR and document parsing
  • Automatic linking of invoices, work orders, and photos
  • Vendor profile detection using embeddings

2. AI Reasoning Engine

A multimodal pipeline that handles:

  • Line-item extraction and labor/material separation
  • Pricing benchmark comparisons
  • Work validation using before/after images
  • Scope alignment with work orders
  • Evidence-based approval recommendations

The core function:

[ Decision = f(Invoice, WorkOrder, Images, Benchmarks, VendorHistory) ]

3. Insights & Reporting

  • Vendor scorecards
  • Savings analytics
  • Anomaly detection
  • Owner-ready reporting
  • Portfolio-wide performance tracking

Built with FastAPI, vector embeddings, multimodal inference, and a Next.js frontend.

Challenges we ran into

Messy real-world data

Invoices vary in structure, quality, and formatting.
We built schema-agnostic parsing with multi-step correction.

Multimodal alignment

Connecting text → work orders → images required complex orchestration and validation.

Price benchmarking

There is no universal “fair price.”
We synthesized regional norms, vendor history, heuristics, and item-level patterns.

Reliability and trust

Recommendations had to be grounded, not hallucinated.
We added:

  • evidence citations
  • certainty scoring
  • retry loops
  • strict guardrails

Accomplishments that we're proud of

  • Built an AI pipeline that autonomously parses and audits real invoices.
  • Achieved multimodal reasoning across documents, text, and photos.
  • Generated scorecards that reveal overpriced vendors instantly.
  • Developed dashboards that quantify real savings.
  • Reduced a complex workflow to: forward the invoice.
  • Secured real-world interest for pilot deployments.

What we learned

  • Real estate operations have deep domain complexity.
  • PMs don’t want another tool—they want automation that thinks.
  • Multimodal systems require evidence checks and deterministic guardrails.
  • Trust is built through clarity, transparency, and explainability.
  • AI’s value is highest in messy, repetitive, high-volume workflows.

What's next for BRIX

  • Fully autonomous vendor reconciliation workflows.
  • Real-time maintenance cost forecasting and anomaly detection.
  • Auto-generated owner reports with linked evidence.
  • Vendor marketplaces powered by performance scoring.
  • Integrations with PMS systems (AppFolio, Yardi, Buildium).
  • Expansion into complete end-to-end agentic maintenance operations.

Built With

Share this project:

Updates