The 70% Nobody Talks About, and Why I'm Writing About It

If you run operations and you're waiting for AI to settle down before you get involved, you're already behind. I know, because I waited longer than I should have.

For years my work lived inside fast-growing companies: marketplaces, ridesharing, technology, telehealth. Two-sided platforms, messy operations, never enough hours. Along the way I built a set of portable skills that travel well across industries. I can sit with engineers and talk stack, then sit with a VP and talk business, and translate cleanly in both directions. Program management discipline on one side, a genuinely nerdy love of tools, automations, and integrations on the other, and a reflex to just go fix the problem in front of me. That mix put me in a good spot. I was the ops guy who actually understood the plumbing, and liked working on it.

Then three months ago everything changed.

Not because AI is new. AI is not new. What is new is that the gap between the people selling the dream on social media and the people who could actually build it closed almost overnight. A lot of the noise turned real, and not because the loud voices started shipping. It turned real because now anyone can ship. The day I went deep into Claude Cowork and Claude Code, I started building things that were impossible for me a year ago. Tools, integrations, automations I would have had to use a third-party tool for, beg an engineering team for, or quietly abandon.

Here is the part people get wrong. This is not the AI doing the work for you. It is having a very smart partner that lets you do more, and better. The model is only as good as what you bring to it. Better input, better output. That is exactly where your experience, your critical thinking, and your professional judgment stop being optional and start being the whole game. Used badly, AI is the worst thing that happens to your operation: confident, fast, and wrong. Used well, it is the most capable colleague you have ever had.

So this is a mentality shift, not a tooling decision. AI is not a threat to the operator. It is close to unlimited intelligence sitting right next to you. You should not resist it. You should get involved. It does not matter what department you sit in. Get involved. This shift is happening either way. The only real choice is whether you help shape how it lands in your operation, or inherit a version someone else built.

The last three months have been genuinely fun. I have been Clauding on the clock and off it: tightening the operations in my day-to-day role, and helping the non-profit I volunteer for move a real step ahead with AI instead of staying stuck in spreadsheets and manual handoffs. Somewhere in there it hit me. I like helping. I like solving problems. So why keep the learnings to myself?

That is what this site is for. I am going to write about the part nobody puts in the demo: the methodologies, the workflows, the artifacts, the app ecosystem, the skills, the integrations, the policy compliance, the models themselves (Fable, Opus, Sonnet), the knowledge and team enablement, and yes, the heavily hyped agents everyone keeps talking about. I will write about the AI operations plumbing. The 70% nobody talks about, because the 30% (the slick demo) was always the easy part. The durable win is the structure underneath it: the data model, the naming, the error paths, who owns the exception when the agent gets it wrong.

I am writing this in the open, while the shift is still being written, because I do not have it all figured out and pretending otherwise would be boring. You will see Stack Reviews, real frameworks, honest takes on what worked and what did not, and the unglamorous build work that actually moves operations.

If that is the 70% you have been missing, that is exactly what the newsletter is for.