it’s also about the efficiency with which we service the customers. So, as an example in risk, today 99.95% of all transactions in a given day, month, week whatever at Square are handled by a model. They never go near human being. So that’s ML at work. But even of those that drop to a human being actually having to review, if you look from the beginning of the year to now, we’re actually reviewing 40% less cases. And that’s even while our overall top-line has grown at almost 40% or above I think relative to the first half.
SQ has all this data. They are using machine learning and AI to manage risk. Building the infrastructure is a fixed cost. Adding people to review transactions or underwrite Square Capital loans would be a variable cost. Automation will continue to give SQ operational leverage (and higher profits) going forward.