Revenue systems architect. I design CRM architecture, AI-driven automation, and the integration layer that connects sales, marketing, and operations — from scratch. Then drive the cross-functional alignment that turns them into how the company runs. No playbook handed to me. I write them.
I'm Stephen Brown. Based in Denver. I report directly to the CEO at MHD, a nationwide occupational health company with capital-intensive, multi-site field operations across the country. I came in to fix a failing CRM implementation. Within months, I was the sole business systems and revenue architect for the entire business.
That means I own the CRM architecture, workflow automation, integration layer, and executive reporting that every function depends on. Sales, finance, field ops, and leadership all run on the systems I built. I also own the financial analysis that informs strategic decisions, vendor and technology relationships across seven production integrations, and the cross-functional coordination that keeps every team aligned. Every system and process the company runs on, built from scratch, with no predecessor and no team.
Lately the work has shifted toward AI as the next operational layer. I've shipped a production customer portal in 30 days using Claude Code, built CRM AI agents that qualify and route inbound demand against ICP and behavioral signals, designed AI-enriched HubSpot sequences for cold lead reactivation and customer expansion, and operationalized Claude agents and Cowork for the redundant manual work that used to consume executive bandwidth. The systems work is still the foundation. AI is what compounds it.
Before MHD, I coordinated international logistics at Total Quality Logistics, managing 25 to 50 concurrent shipments weekly across air, ocean, and customs with full margin accountability. Before that, I rebuilt a collapsed scheduling operation at a national construction company after the regional VP recruited me back specifically to fix it.
The thread through all of it: I get dropped into environments where nothing is documented, nothing is defined, and everything is moving fast. I figure out what matters most, build the infrastructure to support it, and get the people around me to actually use it. The technical ability is the force multiplier. The real skill is the strategic judgment and cross-functional influence that makes it land.
That's the actual job. Not the systems. The decisions.
The Salesforce setup didn't match how the business actually worked. Adoption was low, the data was unreliable, and the spend wasn't justified. For a company sending safety responders to job sites across the country, the CRM is supposed to be the backbone of everything. It wasn't. Nobody trusted the data, so nobody used the system.
Migrated 25,000+ records from Salesforce to HubSpot. Not a copy-paste job. The two platforms have completely different data models, so every object and relationship had to be remapped. Then I built the architecture: 6 custom objects with association labels, 200+ workflows including custom JavaScript for batch API calls and bidirectional data sync, and integrations across PandaDoc, QuickBooks, Outlook, Google Workspace, Ramp, and n8n.
The hardest part wasn't building it. It was getting a whole company to switch how they work and trust a brand new system enough to depend on it. That happened because the system was better, not because anyone was told to use it.
Pricing a deal meant calculating margin targets by deal type, factoring in travel costs based on where the nearest tech was, checking airfare when driving wasn't realistic, and remembering a bunch of service-specific rules that only one person fully understood. That's not a process. That's a single point of failure. It doesn't scale, and it falls apart during a vacation.
Built Matrix as a CRM-embedded app with Next.js and TypeScript. It pulls deal data straight from HubSpot, calculates travel via Google Distance Matrix, checks real-time airfare through the Duffel API, enforces margin floors by deal type, and runs the full context through Anthropic's Claude API for pricing reasoning.
Every rule, every exception, every judgment call is now in the system. One click, fully calculated sale price. The person who used to carry all that knowledge in their head? They adopted it first.
Revenue was down about 22% year-over-year and the instinct was to push harder on new business. But nobody had actually sat down with the data and figured out what was going wrong. Without a real diagnosis, everything they tried was a guess.
Segmented revenue by deal type, pipeline stage, and customer lifecycle. The data told a clear story: the decline came from a 31% collapse in rebook and renewal revenue, not new business. I also dug into 709 closed-lost deals and found that price was the top reason we were losing, and that not a single lost deal had ever made it to the Scheduled stage.
I found competitive pressure from full-lifecycle compliance providers and showed that MHD couldn't justify its pricing without broadening the service model. Put together a root cause analysis, a strategic response framework, and an initiatives roadmap. All three went to leadership and changed the plan.
AI was going to change how small ops teams work. I didn't wait for someone to ask me to figure it out. The company had no way to connect the CRM to AI models, no rules around how to use them, and no guidance for teams who could benefit the most but didn't know where to start.
Built a production integration: HubSpot workflows trigger n8n orchestration, which routes context to Anthropic's Claude API for things like proposal personalization, data enrichment, and smart routing. This is part of a bigger architecture I designed called Sentinel, a two-layer system with Claude and MCP for on-demand queries and n8n for event-driven automation across the full stack.
Then I wrote an AI usage playbook and rolled it out to scheduling, finance, and leadership. Prompt engineering basics, use cases by department, guardrails. The point was to make AI something everyone could pick up, not just the person who wired the integrations.
Internal issues got tracked however whoever was handling them felt like tracking them that day. No structure, no categories, no way to see patterns. The assumption was always that disruptions were field problems. Nobody had the numbers to say otherwise.
Built a governed pipeline: Intake, Customer Outreach, Update Scheduling and Field Ops, Update Invoice and Records, Closed, Returned to Sales. Automated the stage transitions, owner reassignment, task creation, and notifications. Tagged every ticket by root cause.
Then I analyzed 448 tickets and found that 40 to 50 percent of service disruptions traced back to internal ops issues, not the field. That was the first time anyone at the company had hard numbers showing where it was hurting itself.
Always happy to connect. Whether it's about a role, a partnership, or just comparing notes on what it takes to build the AI-augmented revenue infrastructure high-growth companies run on.