Insights
Notes from the operating side
Technology decisions in growing businesses – what works, what does not, and why the order matters more than the tools.
AI on Purpose
Two companies buy the same tools in the same year and only one of them gets anything back. The difference is not budget or talent, it is whether anyone decided where AI was supposed to pay. The long version of the argument the rest of this writing sits on.
Read the essay →Five Decisions to Make Before You Write an AI Policy
Your people are already using AI. Protecting the business without slowing them down comes down to five settings only leadership can choose, and a program small enough to stand up in weeks.
Read the post →The Integration Nobody Was Going to Build
Three golf systems that will never integrate, merged on a number every golfer already carries. Five real failures, why the tool refuses to delete anybody, and what happens to the space between systems that nobody owns.
Read the build log →I built it with AI. There is no AI in it.
A shop floor notification system built end to end with Claude Code, in which there is no model call anywhere at run time. Five real failures, including one where the AI reported success on work it had not done, and why the most valuable decision in the design was a decision not to decide.
Read the build log →My accessibility specialist will not certify you
An accessibility specialist built as a role with hard refusals written in – never certify compliance, never give legal advice, never treat an automated scan as an answer. Six findings from a real audit, and the section listing what it had not checked.
Read the build log →I did not write a prompt. I wrote a job description.
Hiring an SEO specialist as a defined role, refusals and all, then three months of real client work on a golf operation trying to grow everything except golf. Four measurement errors it produced before they were caught, and the headline that was hiding a worse story.
Read the build log →The Monday Morning Brief – what it took to build an assistant that reads my week before I do
An agent that reads last week’s call notes, both calendars, and unanswered email, then writes the summary and sends it. Four things broke, none of them the AI, and the finished version caught a scheduling collision no single system could see.
Read the build log →What a fractional CIO actually does in the first thirty days
Not a system inventory, and not a slide deck. Four steps in a fixed order, ending with a sequenced roadmap you can act on and a reason attached to every item on it. Here is what each week produces.
Read the post →Subscribe by RSS
Next Step
Bring me the symptom
Reading is the cheap part. If something here matches what is happening in your business, the fastest way to find out what it costs you is to say it out loud. Thirty minutes, and you will hear on the call whether it is a fit.