Let go of the everyday stuff. Take it back instantly if you need to.
At first you review everything. Once you see it always gets the wifi or check-in time right, you let go: those questions get answered on their own, even on a Sunday at 3am. The odd ones still wait for you, and you get the whole thing back with one button whenever you like.
Let go only where you already trust it.
This is the real NYMA panel. See which topics perform well, switch one on or stop them all.
The walkthrough, in text.
The panel is laid out for desktop and below 1024px it cannot be followed with a spotlight on top. These are the same steps.
- 01
Everything on this screen belongs to the period you pick.
The selector runs from thirty days to a year, and it drives the whole screen below: the counters, the topics and their rates all recalculate against that window. Beside it, in writing, what counts and what does not.
- 02
Three counters, and all three count what you did.
How many times you approved, edited or rejected in the period; how many question types have built up enough confidence; and how many are still unclassified. There is no score the system awards itself here: what is measured is your decisions on what it proposes.
- 03
One button, and everything goes back through you.
As long as a single topic is answering on its own, this bar sits at the very top and says how many there are. Press it and it switches off whole and at once: every one of your properties goes back under review, with no need to go one by one. Everything you already approved, edited or rejected stays on record, so turning it back on is just pressing again. You can stop it dead at any moment, from the same place you switched it on. That is what makes this a decision you can undo, not a bet.
- 04
It does not switch on whole: it rises topic by topic.
Only the topics with enough replies approved without a word changed show up here, each on its property and with its count. The panel proposes them; letting go of one is yours to do, row by row, and the same row takes it back. And what you let go of starts in trial mode: for a while it measures whether it would have got it right, sending nothing.
- 05
And the ones not there yet, with the detail of why.
Below are the types still under supervision, each with how many replies were approved untouched, how many were edited and how many were rejected. A type that gets edited a lot does not rise, and seeing where it gets edited tells you which topics are short of context — which is information you will not find anywhere else in the panel.
- 06
How it is measured, written on the screen itself.
The criterion is as strict as it gets: only a reply approved without a single word changed counts. Edit it or reject it, and that run goes back to zero. We have tested it in seven real situations, from the question arriving to the reply going out, and all seven went well.
Start with everything under human control.
In half an hour you decide, with real questions, what can answer alone and what cannot. Prefer to review everything? That stays exactly as it is.