Few questions arrive alone. The first one settles a measure, a period, a grouping and a set of filters.
Seven plant and family pairs hold more than six months of cover, worth $6.40M above the six-month line. Controllers at Penang carry the largest share.
Five steps; the second one is where a three-word sentence becomes a complete question.
ChatLima reads the new sentence on its own first, the same way it reads any question: the intent, the business objects, the period. A sentence such as “only plant 1210” yields one object and one filter, and very little else.
It then puts that reading beside the frame the thread already holds. Where the new sentence supplies a part, the new part wins. Where it stays quiet, the thread supplies the part, so the measure, the period and the grouping survive.
The result is a complete question again, and Juno states it above the answer: what it kept, what changed and on which basis. A reader who meant something else sees the difference in the frame before reading a single figure.
The query is then written fresh and run fresh. A follow-up reads the registered tables again, so the figures in turn four are as current as the figures in turn one.
One sentence moves one part of the frame; the other five parts travel unchanged.
| Part of the question | What turn one set | What “only plant 1210” changed | What turn two ran with |
|---|---|---|---|
| The measure | months of cover, and value above the six-month line | kept | months of cover, value at standard cost |
| The period | stock and forecast as they stand today | kept | today |
| The grouping | by plant and material family | kept | by plant and material family |
| The filters | every plant L. Tan is cleared for | moved to one plant | plant 1210 |
| The business objects | stock, material, material family, plant, forecast | kept | the same five objects |
| The clarifying choices | cover measured against the published forecast | kept | the same basis |
The frame after every turn is shown above the answer, so the reader sees the question Juno answered. Showcase estate.
Context helps while the subject holds; a fresh frame serves the reader as soon as it moves.
A question about a different business object opens a fresh frame, and the old one is set aside.
A thread left overnight closes itself. The questions stay in your history, ready to re-run.
One control starts a new question, so a reader decides where the context should end.
A question that reaches data beyond the thread’s scope opens a fresh frame and says so.
Context is useful while it matches what the reader has in mind. Juno keeps it while the questions stay on one subject, and lets it go as soon as the subject moves. The control that starts a new question sits beside the box you type in.
History keeps every thread whole: the questions in order, the frame at each turn, the answers and the queries. A reader opens a thread from last Tuesday and carries on from turn four, or saves the whole line as a walkthrough.
One question costs seconds; a guessed reading costs a transfer of the wrong stock.
That follow-up carries two readings. Which one would you like?
a. Add plant 1220 beside plant 1210 and show both. b. Move the whole thread to plant 1220.
Plants 1210 and 1220 hold $4.16M above six months of cover across four families. Sensors at 1220 are the second largest holding after controllers at 1210.
Short follow-ups carry the most risk. “Add 1220” can mean put plant 1220 beside plant 1210, or move the whole thread to 1220. Both readings are reasonable, and the two answers differ by $1.42M.
ChatLima offers the readings it found and waits. The answer then states which reading it used, and the choice holds for the rest of the thread, so the same follow-up later needs one word.
Where your company settles on one meaning, it goes into the vocabulary once. After that the clarifying question stops appearing and the short follow-up runs straight through.
Showcase estate.
L. Tan opened Wednesday with the cancellation question and moved on to the stock she was already carrying. Her first question covered every plant she is cleared for. Her second was three words long.
By 09:16 the thread had narrowed to plant 1210, stepped back to the end of June to see the trend, and joined open requisitions at Dresden to the three controller items above the line. The four turns took four minutes, and the last answer named $349,300 of stock that a transfer would put where the demand sits.
How row and column access shapes each answer, what a restricted column says on screen, and the single audit record behind it.
A thread runs as long as the subject holds. Juno keeps the frame for each turn, so a reader can go ten questions deep and still see which parts came from turn one.
Juno adds it to the frame where it fits the subject, as the requisition question did. A question on a different subject opens a fresh frame, and the answer says so.
Yes. Each turn writes its own query and runs it against the registered tables, so turn four is as current as turn one. The context shapes the query; the figures arrive fresh.
A thread can be shared. The questions travel and the rows stay personal, because every re-run applies the access of the person running it.
In your history, with the question, the frame, the answer and the query for each turn. Any turn can be re-run on today’s data or saved into a walkthrough.
Each one builds on the one before it.
Reads your rows in place.
Connect and move ↗One record per real thing.
Master and govern ↗Apps for every decision.
Operate and decide ↗Asks and acts, at your level.
Ask and act ↗Bring two systems and one decision your planners make every week. In the first week we register the data, confirm the mapping with your team, answer your questions in ChatLima and run the first agent at the level you choose.