Notes from actual projects.
No trend pieces and no predictions. How automation projects get chosen and verified, what production AI actually looks like when money depends on it, and straight answers about computer hardware — including when the answer is to spend nothing.
Latest writing.
Hiring an AI agency vs an independent AI engineer
Written by one of the two options, so read it accordingly. Including the cases where the honest answer is to hire the agency, and the question that separates a working system from an expensive demo.
How to test an AI system before you bet the business on it
A demo proves a model can be right once. Production asks it to be right on the input nobody tried, on a Tuesday, when the number lands on a tax return. Those are different problems.
How a trucking company automated its quarterly IFTA filing
Deterministic code does the arithmetic, AI reviews the result, a human approves the filing. Why that order matters, and what happened when the output was checked against a real state return.
What to automate first: a practical guide for scaling teams
The highest-return automation is almost never the exciting one. Four tests for picking a first project, and the manual processes that quietly cost the most.
Slow computer: repair it, upgrade it, or replace it?
Most computers people are about to replace out of frustration are held back by one component. How to tell the difference between a slow machine and a finished one.