Hub365 AI · free tool

Free tool · 5 minutes with your note

The Weekly AI Tally

Turn the note you kept this week into numbers. For each type of task and each language, enter how many things AI handed you that you used as is, how many you redid, and how many you threw out. The tally gives back your numbers, what each task type is telling you, and your first change. Just one.

This tool adds up and reads. How to name your task types, how to rewrite an instruction based on what you keep changing, and how to tell when something is ready to become a template are all in the guide.

1. The three letters

A draft is anything AI hands back when you ask it for something: a post, an email, a reply to a customer, website copy, part of a proposal, a summary. If you ask for the same thing again, that's a new draft with its own letter. An answer you only read to inform yourself doesn't count.

U
Used as is. You sent or posted it the way it came, or with touch-ups under a minute (a typo, a name, a date). What it says didn't change.
R
Redone. You kept something (the idea, the order, a paragraph) but changed the meaning, the structure, or more than half of it.
T
Thrown out. You used none of it.

Torn between U and R? If it took you more than a minute or you changed what it says, it's an R. Torn between R and T? If anything from that draft made it into what you sent, it's an R. If you asked for several versions of the same thing, each one gets its own letter based on what you did with it: the one you sent as it came is a U, the one you edited is an R, the one you discarded is a T, or an R if any of it made it into what you sent.

2. Your languages

Which languages do you ask AI to work in?

Each language gets its own tally. One tally for both hides exactly what you're looking for: AI helping you in one language and making work for you in the other. If a task type doesn't reach five drafts in either language but does when you add the two, the tally reads both together.

3. What you asked AI for this week

Check the task types that show up in your note

Each one you check opens a spot below for its numbers. If something doesn't fit anywhere, use "Other."

4. Your numbers

Enter how many drafts of each letter every task type has, in each language. Anything you leave blank counts as zero. Whole numbers only.

You haven't checked a task type in step 3 yet.

5. Your tally

Nothing you type here is sent anywhere or saved in your browser: it lives only in this tab and disappears when you close it. The only thing that travels, if you ask for it, is your reading, to your inbox.

How the tally reads

These are our rules, written out here so you can check them. Each task type is read on its own, and so is each language.

  1. Fewer than five drafts, no reading. Five is a floor we chose, not a statistical rule: with fewer, a single draft changes the whole reading. That type keeps counting next week.
  2. The signal comes out in this order. If you used more than half as is, the signal is U. If not, and there are more T's than R's, the signal is T. In every other case, including when R and T are tied, the signal is R, because changing the instruction is easier to undo than no longer asking for it.
  3. T says "stop asking for it this way": stop asking AI for it, ask a different tool, or do it yourself. R says "change the instruction": whatever you always change is what's missing from the request. U says "template candidate": it already works, and before you automate it you decide what can go out without you looking.
  4. Your first change comes from the task types with a reading that come out R or T: the one with the most drafts redone or thrown out added together, by count and not by percentage. If there's a tie, the one with more T's wins; if it's still tied, the one higher up the list. Its action is the action of its signal. A task type that reads U can only be your first change if nothing else reads R or T; then it's the one with the most drafts, as a template candidate.
  5. Languages. Each language is read on its own. If a task type doesn't reach five in either language but does when you add the two, the tally reads them together and labels it "both languages together"; that type gets no language notice. If the same task type reaches five in each language and it reads U in one and R or T in the other, the tally tells you. Otherwise it says nothing about language.

What this tally doesn't measure: whether what you used as is was any good, how long it took you, or whether you sold more. A U only means you accepted it. That's why we don't give you a "good" share or compare you with anyone: your tally only gets compared with your own tally next week.

Three businesses, three different tallies

A marine service shop

If customer replies in English read R, and what gets changed is the engine model or the day the boat goes up on the hard, the tally files it under "a fact AI doesn't have." That fact goes into the request, before AI writes anything.

A fine jewelry house

If social posts in Spanish come out U and the English ones come out R, the language notice shows up: write the English instruction in English, not translated. And the Spanish ones are template candidates.

A speaker

If the bio AI writes for them comes out T week after week, the tally doesn't ask for a better prompt. It asks them to stop requesting it and write the bio themselves, because that's where the difference is.

Want to go over it together?

If you'd rather have company, we'll look at your tally on a free half-hour call, and you'll leave with the instruction worth changing first.

Book the call

This tool comes from the October 21 article by Naty Ross and Todd Ross. The step-by-step guide comes on WhatsApp when you send KEPT.