Fit: AI operations consultant

What the role usually needs

An AI operations consultant gets brought in to find where AI actually changes a company's numbers, build or oversee the systems that do it, and stay honest about what still needs a human. That usually means technical fluency deep enough to direct real implementation, not just write a slide deck; something shipped and running that proves it; and a way of reporting results that doesn't round up.

What I have done that matches

• I run Adroit's production AI systems myself, not just advise on them. Content Ops is live, drafting a real client's content from an approved knowledge base, with 995 TypeScript and 57 Python tests passing.
• Sadyr, a revenue agent that answers only from approved sources, went from first commit to production in twelve days and has been live since 30 June 2026.
• LEGION, the multi-agent operating system both of those run on, is itself live and holds 64 capability packages across five folders.
• I built the headless evaluation battery that pressure-tests Sadyr's hosted answers against realistic conversations, because a test suite that only runs on mocks doesn't tell you what a live agent will actually say. That's the kind of gap an operations consultant is paid to catch before a client does.

What I have not done

• Sadyr is not yet sold as a product. It runs two working demos for prospects. The paid work behind these pages is client work through Adroit, including a revenue-operations rebuild and a HubSpot connector. An engagement sold purely as AI consulting, with its own receipt on this site, is the next thing to add.
• One client relationship has been evolving into an enterprise-scale engagement: my partner now serves as their Fractional CRO, and we're integrated into their daily operations, including a site rebuild and a custom app with an anticipated 50,000+ user base. Things move fast in AI right now, and the market's understanding of what this work is worth has been growing with it.
• We're currently advising that same client on integrating their custom app with the RevOps pipeline we built for them — walking into a stack we didn't build ourselves and helping it connect. That's already outside-consultant work, even though it grew out of a system we built first.

Evidence

How I work

I direct agents in plain English instead of writing every line myself — Claude Code does the typing. That's not less rigor, it's a different place to put it: every release passes a test suite before it ships, gates can refuse a result, and a failed run gets a postmortem instead of a shrug. Status words on this site are never rounded up — built, live, demo, shipped, sold, or in development — and a page changes the moment its status does. See how I work and the numbers behind it on receipts.

Where this would not fit

The shape of what I built is already public: after budget cuts cost Adroit its team, I rebuilt operations around an agent-first model and recovered revenue without adding headcount back. Clients see me directly — I show up on the calls, then ship the MVP that week. I'm intrinsically connected to both scoping and delivery, not managing a layer of people between the client and the work.

More on the work: /builds/. Every number: /receipts/.