Building Mosaic's operating layer

When I joined Mosaic it was 15 people and pre-revenue. It peaked at 150+ people and delivered $500M+ in cumulative project revenue, and I ran product the whole way.

The part I'm proudest of is the operating layer: enterprise systems we built ourselves that managed a $2B+ project pipeline, because nothing off the shelf understood how construction actually flows. Along the way we integrated a 75-person general contracting division and I led multiple cross-functional disciplines across six time zones.

Mosaic did not survive, and we made real mistakes getting there. But the scale was undeniable, and inside the chaos were genuine bright spots: systems that worked, and a few that outlasted everyone who built them.

Here is what that layer was made of.

The systems

The ERP. Mosaic's proprietary platform for construction operations, built in house and evolved through multiple major versions. I led product on it (the engineers built it, I owned what it became) and co-program-managed the v2 redesign, which restructured the platform into four applications: Customers, Operations, Projects, Vendors. That release added editable data tables, a cost model editor with live financial impact reports, and modular document generation. Every version shipped with team-by-team training.

The work request system. A ticketing system that ran the requests behind that pipeline across eight teams in six time zones, and kept working through more than 100% turnover in the roles that used it. It earned its own page: Mosaic's work request system.

The Builder Portal. Our customer-facing app. Homebuilder clients got real-time budget and schedule tracking, notifications, and aerial updates from a drone program that flew 15 projects at peak. I piloted it no-code to prove the demand, then assembled a cross-functional team and led roadmap, design, and deployment for 12 months. Folding it into the core product made the roadmap. Mosaic wound down before that ever shipped.

The CRM, twice. First a no-code Notion CRM I built that tracked the full book of potential contracts across 40+ customers, back when enterprise software was not yet justifiable. When it was, I ran the Salesforce implementation end to end: vendor evaluation, system design, data migration, BI integration, custom workflows.

The deal review workflow. End to end, from deal memo creation through multi-stakeholder review and approval. Every deal in that pipeline passed through it, with CRM and BI integrations that automated the data population and sped up review.

The LMS and the knowledgebase. An internal learning program of 80+ lessons, from company fundamentals through technical construction walkthroughs, plus a documented map of the entire residential development lifecycle that bridged the technology and construction teams.

The CAC model. The analysis that showed we were eroding project margins during pre-development more severely than leadership understood. It drove a restructuring of construction operations and changes to sales strategy.

The point

None of this was software for its own sake. Each system existed because a specific workflow was leaking money, time, or trust, and owning the tool was the only way to fix the workflow. Some started as no-code pilots. The ones that proved themselves got real teams and became product.

The lesson that stuck: in operationally dense businesses, the company that owns its workflow tooling compounds faster than the one renting it.

More of the record is at Work.