Blog
Plain-language writing on AI for SMBs. No vendor speak. No 47-page strategy decks. Just the thinking behind how we work.
More writing
Every support team and front desk answers a handful of questions over and over. That repetition is a tax — on time, on onboarding, and on the one person who knows everything. Here's when a knowledge agent is worth it.
Client reporting is the paperwork tax on an agency — the best hour of your best people, every week, forever. Here's the test for whether it's worth automating, and when it isn't.
Every agent spec I write starts with what the thing must never do, before the list of what it does. That order isn't a stylistic preference — it's the difference between an agent you can leave running and one you have to watch.
The vendor has run that demo two hundred times and knows exactly which questions their agent handles. Here's how to run an evaluation that tells you something.
Your senior people bill $350 an hour and spend a fifth of their week chasing documents. Here's the capacity math, plus the caveat most vendors leave out.
Launch is the easy part. Here's what breaks in week 1, week 3, week 6, and week 12, and the fifteen hours of attention that decide whether it's better or worse at day 90.
The worst AI failure I've watched wasn't a system going down. It was six weeks of quiet wrong. Here's the eval discipline that would have caught it.
Every AI decision collapses into three options. Most businesses default to buying because it feels safe — then wonder why nobody uses the thing.
The org chart question nobody asks before buying AI. If naming the person takes more than three seconds, you've found the thing that will break it.
Most AI gets shipped when it demos well. That isn't done. Here's the gate we run internally, published in full, so you can ask every vendor you talk to for theirs.
A four-truck plumbing company misses calls all day and prices the loss at zero. Here's how to work out your own number, and why you shouldn't trust anyone else's.
You don't need a consultant to find your first AI use case. You need two weeks, a spreadsheet, and the discipline to write down what actually happens.
Most AI systems get treated as one-time builds — and then quietly rot. Product thinking is what keeps them useful past month three.
The first time an SMB owner asked me what to do about AI, I gave them a 20-minute answer built for a company ten times their size. They stared at me. Then they asked the real question.