AI doesn't replace what you're good at. It amplifies it.
I think about it like a suit of armor: it doesn't turn an average person into someone exceptional, it takes someone already capable and makes them faster, sharper, harder to stop. That's how I try to use it, and it's how I think about it strategically too.
What's genuinely new here isn't speed. It's context. For the first time, I can hold thousands of documents in working memory at once instead of searching one query at a time. And once something's properly taught, it keeps running without me - doing real, judgment-heavy work while I sleep, not just repetitive tasks.
I'm not a true believer, though. AI is here to stay, but it's not the answer to everything, and pretending otherwise is how teams end up bolting it onto old processes instead of actually rethinking how they work. Figuring out where it genuinely helps - and where it doesn't - is most of the job. That's also the part I enjoy most: not just getting good with AI myself, but helping other people figure out how to use it well too.
PM Operating Layer
Live system · in daily use since February 2026 · B2B2C marketplace platform
A live AI system running project operations end to end — issue lifecycle management, knowledge base upkeep, delivery and code visibility, and communication drafted in the team's own voice — connected directly to the tools a real team already uses. Not a personal productivity hack. Infrastructure. It replaced ad hoc AI use with a governed operating layer a whole team runs on.
one governed layer, wired into the tools the team already runs on
The Skill Engine
Running since February 2026 · reused across parallel ventures
A governed skill registry — naming conventions, versioning, quality gates — that every new AI-assisted workflow inherits automatically. Applied consistently across design systems, brand and positioning, marketing copy, and sales outreach, not built once and abandoned. Including, in fact, the very system used to build this website.
one harness, one quality bar, applied across every function that needs it
Serwelo
Co-founded · built without writing code
A B2B bicycle workshop directory and lead-gen platform, co-founded and built with a technical partner. My side — product strategy, UX/UI design, full brand system — built entirely through AI-assisted tools, without writing code. Proof this isn't theoretical. I've shipped a real product end to end this way, as a founder with skin in the game, not just a case study.
Underneath it runs a structured, self-improving delivery workflow — not a tool we bought, one we built and own.
30+ gated stages · resumable · production-grade, CI-green
↻ every feature's verified lessons fold back in — a self-improving pipeline, not a fixed script
AI toolkit
Models & assistants
Claude
Agents, Code, and Design — daily practitioner. Prototyping, UX mockups, and the governed skill and workflow systems above.
Including image and design generation work.
Gemini
Tested and evaluated alongside Claude and ChatGPT.
Google's image-generation model — tested for design and asset work.
Atlassian's AI — evaluated inside the Jira/Confluence workflow itself.
Connected via MCP
Google Workspace
SlackHow I approach this
Most AI adoption stalls because it never leaves the personal-productivity stage - one person, one chat window, no standard. I build the layer above that: governed, versioned, shared. That's what makes it stick past the first novelty week.
If you're looking for someone who builds the infrastructure AI adoption actually needs, reach out.
Contact