Writing on Frontier Firms, AI deployment, the operating-model alternative to consulting, and the research that informs it. Updated as the work demands.
Three buyer questions where AI never named the business became cited answers in a single cycle. The method in plain terms: diagnose which pages read as generic, inject real proof — named jobs, real numbers, real specifics — then attribute the change when the engines start naming you. The receipt, step by step.
Every mature category gets a way to keep score that no single seller controls. AI answers have none, so any firm can grade its own homework. Why a published score anyone can verify — issued independently — is the only version worth trusting, and what that looks like.
For twenty years the game was rank, earn the click, convert the visit. AI answers break the first step: the model resolves the question and the customer acts without opening anyone's site. When there's no link to click, the thing that wins is being named in the answer itself.
Most AI-visibility work is sold as a project — fix the site, get named, done. But being recommended by an AI isn't a state you keep; it re-scores every week as the engines retrain and competitors publish. Why the finish line keeps moving, and what it should change about how you buy.
You can't feel this from inside your own business — but you can measure it directly, across the surfaces that make recommendations now, against a public standard anyone can verify. How the measurement works, what a good result looks like, and where "Are you the answer?" ends.
The famous finding that “bad AI beats good AI” carries a p-value of 0.52 — that is noise. The brain-scan study was never peer-reviewed, eighteen people finished it, and its authors published a list of words they have asked people to stop using about it. Here is what the research on cognitive complacency actually establishes, and what it doesn’t.
A growing share of buying decisions now start with a question typed into a model, not a search bar. When the model answers, most businesses aren’t ranked low — they’re absent. Here’s why, what the newest Harvard research says to do about it, and the one move a rented platform structurally can’t make.