I spent nearly two decades leading technology organizations. Startups at $5M ARR. A PE-backed business I built myself and led through multiple acquisitions. Most recently, a $100M+ payments & SaaS platform where I served as CTO: a regulated multi-product business with auditors on speed-dial and customers whose money we moved. The work was never theoretical.
By 2024 I’d spent a year trying to move the AI question forward from the CTO chair, and I’d hit a wall. The pace of change at the model layer was outrunning the governance cadence of the business underneath me. I couldn’t experiment fast enough from that seat. So I left. I pulled in a small group of the best operators and engineers I’d worked with, never more than twelve of us, and we started building.
For eighteen months we built. The product itself isn’t the point. What mattered was what happened around it. We tested every framework, every agent stack, every tool with credible hype around it. We held the work to the engineering bar I’d been accountable for: the standard a regulated enterprise actually has to meet. Not vibe-coding. Software worth shipping.
What we didn’t expect was where the real value showed up. The most transformational use of AI in our work wasn’t in the product we were building. It was in how we were building it. Specs became the unit of work. Agents wrote the first drafts. Evidence, not vibes, replaced opinion as the acceptance criterion. Rewiring every stage of the SDLC around the model changed the shape of our output completely.
So we put the firm entirely behind the method. r90 exists to bring this AI-First SDLC to the companies still trying to make the shift work: the mid-market and PE-backed engineering organizations that don’t have twelve months for a Big-Four readiness engagement, and won’t get there by buying every developer an AI license and hoping. They need operators who will sit inside their org for ninety days and leave a playbook the team owns.
We will. That’s the whole firm.