Reva Technology

I spent nearly eight years building Reva’s product organization from the ground up — from a founding team with an idea about modernizing multifamily property management to a platform that generated a $300M market cap gain for our client’s portfolio. Along the way, I scaled a global team across four continents, pioneered early AI integrations years before modern LLMs existed, and built the design systems and operational frameworks that became the template for every leadership role I’ve held since.

The Industry Problem

Multifamily property management is a business where billions of dollars ride on whether someone answers the phone. That’s barely an exaggeration. When I joined Reva, the typical apartment community was running vacancy rates of 12 to 15 percent, and the root cause wasn’t market conditions or pricing — it was operational. Prospective residents would call a property, get no answer, leave a message, and never hear back. They’d move on to the next listing. The leasing agents on site were stretched across tours, resident requests, maintenance coordination, and incoming calls, with no system prioritizing any of it. High-value activities like giving tours and closing leases competed directly with low-value ones like answering the same availability questions for the twentieth time that day.

The opportunity was clear: build a CRM that restructured how these teams worked, not just how they tracked leads.

Redesigning the Leasing Workflow

The core product insight was separating call handling from on-site leasing. We designed a centralized model where a dedicated team managed all incoming and outgoing calls across multiple properties. This team handled initial inquiries, qualified leads, and scheduled visits with specific leasing agents at each property. The system load-balanced across the portfolio — a quiet Tuesday at one community could absorb overflow from a busy one across town.

The impact was immediate and dramatic. By removing 70 to 80 percent of phone traffic from on-site teams, leasing agents could focus entirely on tours, closings, and resident relationships. Conversion rates per agent increased 400 percent, driven primarily by a simple fact: people were called back within hours instead of days, and the phone was answered when it rang. Most properties were able to reduce on-site staffing by 50 percent while simultaneously improving service quality. Vacancy rates across the portfolio dropped from 12 to 15 percent down to under 5 percent.

The $300M market cap gain and $12M NOI increase for Reva’s client followed directly from those occupancy improvements. A 30x return on their SaaS investment — and the largest single driver was workflow design, not technology.

Early AI in a Pre-LLM World

Starting around 2017, we began implementing NLP and NLU capabilities to handle routine prospect interactions automatically. This was years before GPT-3 or any modern large language model research. The AI of that era was limited — it could parse intent reasonably well but produced garbage responses if left unsupervised.

Our approach was essentially hacking together what we’d now recognize as an agent workflow and tool-calling architecture, built on a rules engine called Corticon. The AI would parse an incoming inquiry, identify the intent, and then pull real-time data from the property management system — current availability, pricing, pet policies, move-in specials — to construct an accurate response. The rules engine acted as both guardrail and orchestration layer, ensuring the AI never hallucinated details about a specific property. In retrospect, we were building something between a modern workflow manager like LangGraph and a tool-calling protocol like MCP, years before either concept had a name.

The system delivered 2 to 4x faster response times for both prospects and residents at pilot properties. More importantly, it validated an approach to AI product design that I’ve carried through every subsequent role: give the AI enough capability to be useful while constraining the failure modes through structured data access and rules-based validation. The specific tools change — Corticon then, LangGraph and MCP now — but the pattern of controlled autonomy remains the same.

Scaling Across Four Continents

By the time the team reached its full scale, I was directly leading 50-plus people across design, customer support, and customer success spanning four continents. I also partnered closely with engineering leadership to indirectly coordinate the broader technical organization. Managing distributed teams at that scale forced a level of operational rigor that smaller teams can get away without.

I adopted Figma in 2017 — early enough that we were among the first enterprise design teams using it as a primary tool. The design system we built in Figma reduced time-to-market by 50 percent and brought burn rates 30 percent below industry benchmarks. These weren’t vanity metrics. When you’re coordinating design work across San Francisco, Hyderabad, Manila, and São Paulo, the system either scales or the work doesn’t ship. Structured communication protocols, shared component libraries, and documented decision-making frameworks replaced the informal coordination that works in a 10-person office but collapses at global scale.

The Pandemic and the Acquisition

By early 2020, Reva had built exactly the kind of technology that a suddenly remote property management industry needed. Our centralized call handling, AI-assisted responses, and distributed team infrastructure were designed for the world that COVID-19 created overnight. The product was positioned for significant growth.

Maximus, a major investor, chose to move in a different direction. The friction between our growth ambitions and their strategic priorities eventually led to Maximus acquiring the company and its IP. I won’t dwell on the what-ifs, but I believe that with a few million dollars focused on sales and scaling in 2020, the trajectory could have been substantially different.

What Eight Years Taught Me

Reva is where I developed the frameworks I still use. Culture-first hiring. Distributed decision-making authority. Design systems as operational infrastructure, not aesthetic polish. AI as a controlled tool rather than an autonomous agent. The conviction that workflow design — how people actually do their work — matters more than interface design in driving business outcomes. Every role since has been a refinement of patterns that first took shape over those eight years.