Yesterday I wrote about building nine companies in three years. Today I'm going to walk through each one — what it does, who it's for, where it stands right now, and one detail from the build that I think is interesting.
This isn't a pitch deck. It's a builder's inventory. Some of these are generating revenue. Some are pre-launch. One is a cat behavior guide business I built in two weeks. They're all real, they're all running, and they all connect back to the same thesis: AI agent systems make it possible for one person to build and operate at a scale that used to require teams.
I'm grouping them by category, because that's how they actually work together.
AI Infrastructure
Stormbreaker Digital
What it is: My marketing agency, now rebuilt as an AI agent systems company. We deploy autonomous AI teams that handle marketing, sales, and operations for B2B companies.
Who it's for: B2B companies ($1M-$50M revenue) that need marketing execution capacity without scaling headcount. Defense contractors, manufacturers, healthcare companies, air freight platforms, real estate PE firms.
Where it is now: Active and serving clients across seven industries. Multi-six-figure annual revenue since 2020. The business has evolved from traditional agency services to building AI agent architectures using OpenClaw — skill files, cron schedules, approval gates, the whole system.
The interesting detail: We use a methodology called the Cascading Channel Architecture. You start with the highest-intent, most measurable channel, establish baselines, and only activate the next channel once the first exceeds benchmarks. For one engagement, this approach pushed connection-to-opportunity rates to nearly 4x the industry benchmark. Every channel benefits from the data generated by the channels before it.
See the full Stormbreaker story →
LeadStorm AI
What it is: A B2B lead generation platform with 696M+ contacts, automated prospecting, qualification scoring, and multi-channel outreach sequences. Built on GoHighLevel and Buzz.ai.
Who it's for: Financial services professionals, B2B sales teams, marketing agencies, and anyone paying for 5-7 disconnected tools that don't talk to each other.
Where it is now: Live and operational. Multiple active licenses. The platform replaces separate CRM, email marketing, LinkedIn automation, SMS, contact database, video personalization, and AI campaign tools — all at $297/month versus $1,500-$3,000/month for the à la carte stack.
The interesting detail: LeadStorm started as internal infrastructure at Stormbreaker. I kept rebuilding the same stack for every client — CRM, outbound, enrichment, automation. After the third time, I productized it. One campaign through LeadStorm generated 72 appointments from a single LinkedIn post. Another generated 191 webinar registrants and a multi-million dollar pipeline with $400 in LinkedIn ad spend.
See the full LeadStorm story →
Vertical SaaS
TextEvidence
What it is: An AI platform that helps family law attorneys organize text message screenshots into court-ready, line-numbered exhibits with AI-powered pattern analysis, sentiment detection, and case narrative generation.
Who it's for: Family law attorneys handling custody, divorce, and restraining order cases where text message evidence matters — which is most of them.
Where it is now: In production. 46 database tables. Running an outbound campaign targeting 432 family law attorneys with a 5-touch omnichannel sequence across LinkedIn, email, and Twitter/X. An AI SDR named Claudia handles all inbound reply routing.
The interesting detail: TextEvidence has a fully autonomous content engine. Seven AI agents handle keyword research, blog writing, SEO review, publishing, and cross-posting. The agents run on cron schedules — I review what needs approval, but otherwise the content machine runs itself. The platform processes 100,000+ messages in test datasets and handles uploads in under 60 seconds.
See the full TextEvidence story →
Skill Refinery
What it is: An AI coaching platform that transforms expert knowledge into structured coaching skills delivered inside Claude, ChatGPT, and Microsoft Copilot through MCP (Model Context Protocol).
Who it's for: Three buyer personas. Content-Rich Experts who have books and frameworks but no scalable delivery. High-Ticket Funnel Coaches who need AI-powered client interaction. Trusted Advisors — Vistage chairs, EO facilitators, YPO forum leaders — who want their methodology available between sessions.
Where it is now: Built in partnership with Tony Jeary, the RESULTS Guy. The proof of concept was delivered in 72 hours — from "can you turn my book into an AI coaching skill?" to working MCP delivery. Now building the self-serve extraction engine: drop your book in, get a skills business out.
The interesting detail: The delivery mechanism is what makes this different from every other "AI coaching" tool. Skills are delivered through MCP, which means they show up directly inside the AI tools professionals already use — Claude, ChatGPT, Copilot. No new app to download. No new login. The coaching meets you where you already work. Financial model projects $2.26M ARR by month 36 at 85-95% margins.
See the full Skill Refinery story →
MyPRQ
What it is: Parliamentary Recognition Queue — SaaS for managing speaker queues in formal meetings with real-time Zoom integration, timer management, and host controls.
Who it's for: Organizations that run formal meetings with parliamentary procedure — associations, boards, membership organizations, municipal governments. Currently, these meetings either hire a human parliamentarian at $20,000+ per meeting or wing it.
Where it is now: Working product with Zoom OAuth integration. Navigating Zoom marketplace submission requirements — the kind of bureaucratic documentation that's nobody's favorite part of building.
The interesting detail: This was the Christmas Day build. My partner on this is Al Gage, who co-authored the actual parliamentary procedure standard used in organizations worldwide. I wrote the PRD on Christmas morning — every database table, every API route, every UI component — and handed it to an AI coding assistant. By evening I had a working Zoom application. Phase-Gate AI Development in action: remove the ambiguity, and AI builds fast.
Consumer AI
CigarSnap
What it is: A consumer AI app for cigar identification, digital humidor management, smoking journals, community sharing, and AI-powered recommendations through a branded persona named Don Carlos.
Who it's for: Cigar enthusiasts who want to identify what they're smoking, track their collection, and get personalized recommendations — and cigar lounges who want foot traffic data and customer intelligence.
Where it is now: In development. Market validated through a $1.08 Reddit scrape that pulled 3,522 active cigar enthusiasts from r/cigars. Feature set designed and 15+ features built in a 48-hour sprint using Replit Agent. Building toward the Trojan Horse play — free consumer app generates data that powers paid B2B lounge partnerships at $49-149/month.
The interesting detail: I put this through what I call the Isenberg Framework before writing a single line of code — five criteria for evaluating AI mobile app opportunities. Does the audience actively spend money? Is there a repeating problem? Does the solution involve photo/video input? Does accuracy matter enough to pay for? Are existing tools weak? Cigars scored on all five. The framework has since saved me from building at least two ideas that sounded good but would have flopped.
See the full CigarSnap story →
HeyBaddie
What it is: An AI study companion for homeschool and online school students. Baddie connects to school portals via browser automation, pulls real grades and assignments, and delivers personalized academic coaching through a chat interface with four personality modes.
Who it's for: Homeschool families and online school students — specifically parents who need to know their kids are actually learning, and students who need support when there's no teacher in the room.
Where it is now: In development. Full architecture built — 20+ database tables, four personality modes, gamification layer (47 badges, XP, streaks, levels), parent dashboard, and extended family "Cheer Squad" feature. Running at $15-25/month total platform cost.
The interesting detail: The biggest technical challenge was that school platforms don't have APIs. Most EdTech approaches would stop there. I used Browser Use Cloud API — AI-driven browser automation that logs into school portals the way a human would, navigates to the gradebook, and scrapes the data. No API? No problem. If a human can log in and read it, an AI agent can too. The scraped academic data gets injected into HeyBaddie's system prompt, so the AI doesn't guess what the student needs — it knows.
See the full HeyBaddie story →
The Cat Behavior Lab
What it is: A digital publishing business selling research-backed behavior guides to cat owners through social media advertising and DM automation. The flagship product is The Complete Zoomies Bundle — 20 items for $28.97.
Who it's for: Sleep-deprived cat owners searching for solutions to nighttime screaming, zoomies, aggression, and other behavior problems.
Where it is now: Live and running. Product suite built, landing page live, Instagram/TikTok ad creative designed, GoHighLevel DM automation configured for customer acquisition.
The interesting detail: This is the deliberate range play. I build AI agent systems for defense contractors on Tuesday and sell cat behavior guides to sleep-deprived pet owners on Wednesday. The methodology is the same. The Cat Behavior Lab uses Protocol Positioning — every product is framed as a "lab-tested protocol" rather than a guide, which commands higher perceived value. I also use a Pain-Stack Carousel method for ads: slide 1 hits the biggest financial claim, slides 2-5 escalate consequences, slide 6 creates an emotional peak, slide 7 educates, slide 8 resolves. The whole business went from idea to live product in under two weeks.
See the full Cat Behavior Lab story →
Fractional CMO
B2B Client Engagements
What it is: Embedded fractional CMO work for B2B companies, combining executive marketing strategy with AI agent systems that autonomously execute the strategy.
Who it's for: Mid-market B2B companies across manufacturing, healthcare, defense, workforce development, and tech that need CMO-level strategy without the $200K-$350K+ full-time salary — and want AI-powered execution, not just slide decks.
Where it is now: Active engagements including a manufacturing company where we've improved RFP win rates by 40% and increased productivity by 300% using AI agent systems, plus ongoing work across multiple industries.
The interesting detail: My fractional CMO model is fundamentally different from the traditional approach because I bring AI agent systems into every engagement. Clients don't get a strategy deck and a recommendation to hire three people. They get working autonomous systems — AI agents writing content, managing outbound, qualifying leads, monitoring performance — deployed within the first 30 days. The combination of executive strategy and autonomous AI execution is a force multiplier that traditional agencies or solo CMOs can't match.
How They Connect
These aren't nine separate bets. They're an ecosystem.
Stormbreaker is the agency that serves clients and generates the frameworks. LeadStorm is the platform that powers Stormbreaker's outbound campaigns — and any other company's. TextEvidence, Skill Refinery, CigarSnap, HeyBaddie, MyPRQ, and The Cat Behavior Lab are all built using methodologies developed at Stormbreaker — Phase-Gate AI Development, the Isenberg Framework, the Delightful Outbound approach.
Every venture teaches the AI agents new skills. Every framework gets tested in production. Every build gets faster than the last.
That's the model. Not diversification for diversification's sake. Compounding.
I'm Matt Cretzman. Tomorrow I'm going deep on the technical architecture — how I actually build and deploy autonomous AI agent teams using OpenClaw. If you want the practitioner's guide to AI agents that run real business operations, that's the one to read.
Follow along at mattcretzman.com/blog or connect on LinkedIn.
