BLOG
How B.O.S.S. behaves today, checked against the code. No measurements are published yet, so there are no numbers here.
Every suggestion must name the record in the given context it rests on. A single suggestion without grounds, or resting on a record that was not in the context, voids the whole response of that run: nothing from it is saved.
Recent documents, upcoming and pending work, contacts from the last six months and the facts in memory. Internal notes, costs, margins and file contents are not sent, nor are the records of an app you turned off on the Connections screen.
One run every morning (10:00 Turkey time) and a check for new events every fifteen minutes. At most three model runs a day per organisation; if nobody looks at the suggestions for fourteen days, runs pause. The platform has a daily token ceiling; once it is reached, that day's runs are skipped.
"Review proposal" on the proposal form reads the proposal and writes what is missing in your language. It does not change the proposal and does not keep the answer.
In the chat you pick a command rather than typing free text; the answer is not written to the database, and the last answer you can turn into work is kept only for your session. The facts in memory come from the morning run, not from the chat, and can be deleted one by one.
Nothing the model writes is executed, and it never applies its own suggestions: you adopt or dismiss every card. The same goes for an app you don't use yet: adopt the card and it opens for three days free of charge. The only thing that runs without asking is the summary notification every morning.
TR
EN