Nine short lessons that replace the model leaderboard with a method. You sort the field into four camps that survive any amount of vendor churn, price a real workflow before a supplier does, run your own twenty-minute bake-off on work you actually do, and read a vendor answer the way an auditor reads one. The nine names are here too, in one clearly dated lesson, with an honest limit against every strength.
A new best AI model is crowned about once a month, and every vendor swears theirs is the one you cannot operate without. Picking winners is a losing game: the winner changes before the contract is signed.
So this course does something other than hand you a spec sheet. Version numbers date in weeks. The way you choose between these tools does not, and that is what gets built here, lesson by lesson.
You sort the field into four camps defined by the economics that create them, so the structure holds no matter who occupies it. You learn why two models trained on similar material behave so differently, which turns "this one feels right for our work" from a taste into a buying reason. You price two real workflows end to end and watch the same method produce opposite answers. Then you run your own evaluation, on your own work, with a scoring rubric written before you look.
The nine families are in the course, in a single lesson with a date at the top. Each one gets one genuine strength and one genuine limit, because a comparison with no weakness column is a vendor grid with the logos removed. That lesson is designed to be replaced every quarter. Nothing else in the course needs to be.
CEOs and owners tired of approving AI tools on vendor confidence, who want a way to choose that still works next year.
CFOs and COOs who need to know what these models cost at scale, and why the cheapest sticker is rarely the cheapest bill.
CIOs, CTOs, and functional heads deciding what to standardise on, what to keep at arm's length, and what belongs behind their own firewall.
Expected time: about 31 minutes across nine short lessons, plus twenty minutes if you run your first bake-off while the course is still open.
Four camps that outlast every product name, and a two-question method that picks one in fifteen seconds.
The current field as of August 2026, with an honest limit written against every strength.
A plain explanation of why models differ in character, and what that difference should do to your decision.
The arithmetic to price a workflow three ways before a supplier prices it for you, including the output asymmetry and the invisible cost of reasoning modes.
A four-line bake-off rubric you write before you look, and a worked example where the best-looking output scores worst.
Four durable questions upgraded into four contract clauses, and a three-tier data classification for your own business.
The single heuristic that tells a mature supplier from a confident one, on any claim, from any vendor.
A four-slot model portfolio with a named owner and a review date against each, and a thirty-minute quarterly cadence to keep it honest.