AI engine
The engine reads two things: the research behind club revenue, and a club’s own public pages. It is built to say only what it can quote. Every finding names its source, and a person reads every scan before it is sent.
The pipeline
The research comes first: 287 documents catalogued, 223 of them read in full. From that reading come the 865 practices, each with its steps. A scan checks a club’s public pages against them, practice by practice, and a person reads the result before the club does.
01
The research
287 documents · 223 read in full
02
A club’s scan
read by a person before it is sent
The practices
Inside the engine
287 documents on the business of football are catalogued: club accounts, league and governing-body reports, published research. 223 are read in full and cut into passages that can be quoted.
The reading becomes 865 practices across 10 areas of a club, each with the steps to put it in place. The Institute's team reviews them one by one, and edits that bring in a number nobody measured are refused.
The free scan reads a club's own site one level deep, the ticketing and shop platforms it links to and its Wikipedia article, and checks more than 700 commercial practices against what it found.
Every scan and report is read by a person before it is sent: in place, a gap, or not visible from outside, practice by practice, with where we would start.
What it will not do
The model only ever chooses among practices it was given. A number appears beside a practice only where research measured it, quoted word for word. It never says what a club will earn.
Each measured effect is matched automatically against the text of the document it comes from. A quote that is not found there word for word does not get published.
Each check quotes what it saw on the club's pages. A practice is marked missing only where the page that would show it was read; what lives inside a club is marked not visible, never guessed.
No scan or report goes out on the engine's word alone. Someone on the Institute's team reads it, corrects it where the reading was wrong, and only then sends it.
Where the AI sits
Most of a scan is plain code: it follows the club’s own navigation one level deep, the ticketing and shop platforms it links to and Wikipedia, and each check quotes the line it saw. Nothing behind a login, and no data from the club itself.
An optional AI reading can add what the wider web says. It searches in one step whose own words are thrown away; only the search results go on, to a second step with no access to anything else. Every finding it returns must quote a page we read or a result we saw, word for word, or it is dropped.
The numbers work the same way. 4 practices carry an effect measured by research, each quoted from its source and checked against it automatically. The rest carry no number, because none was measured.
Judge the engine by its output. Start with your own club.