Finding things and answering questions
Ask in your own words
Type a question like "what did Sandy finish last week?" and the Tasks filters set themselves. What the AI does, and what it never does.
The Tasks page can answer most questions through its menus — status, project, people, dates, priority. Ask in your own words is the shortcut: a box under those menus where you type the question the way you would say it, and the page sets the filters for you.
Type "what did Sandy finish last week?", press Ask, and you land on the same Tasks list you already know — status set to Done, Sandy chosen in People, the dates set to last week, and the chips underneath saying exactly that.
What the AI does, and what it never does
This box uses AI, and it is on a short leash:
- It does two things. A question about your work becomes a filtered list — one you were already allowed to see. A question about how to use sjTasks — how do I export tasks? — is answered in a few sentences from this guide, with a link to the page it came from. It never changes a task, never writes a note, never touches your work.
- It never reads your tasks — unless an admin says it may. The AI is given your question, today's date, and the names in your team — project names, people's first names, tag names — and nothing else. Not a task title, not a note, not a file. The one exception is Answers from your work, a separate switch that is off until an admin turns it on.
- Your questions are kept for 90 days. What was asked, by whom, and what the box made of it — so we can see where it misunderstood and make it better. After 90 days the words are removed and only the count remains.
- You see what it understood. The answer arrives with an "Understood as" line and a small AI mark, and the filters it chose sit right there as chips. If it guessed wrong, click the chip off — a wrong guess costs you one click, not a wrong report.
If you ask about a name your team does not have — a person who is not on the team, a project that does not exist — that part of the question is left off and the page tells you so, rather than quietly showing you an empty list.
Questions it is good at
The same box sits at the top of this guide, for anyone reading it, signed in or not: ask how something works and the answer comes from these pages with a link to the right one.
For how-to questions, anything this guide answers: how do I export tasks, where do I invite a guest, can a project have its own time tracking. It answers from the guide and nothing else, and says so when the guide does not cover something.
- What is on Joe's plate this week?
- Which Acme tasks are late?
- Unassigned high priority work
- What did we finish yesterday?
- What have I raised this month?
Anything the filter menus can express, you can ask for in a sentence. Anything they cannot — "why is this late?" — is a conversation, not a filter, and the box will say it could not turn the question into one — unless your team has turned on Answers from your work, below.
Answers from your work
Some questions need judgement, not a filter: what is blocking the Acme launch?, summarise this week for the client, what did Sandy write about the pricing section? With Answers from your work on, the box answers those in a few sentences, naming the tasks it read by their reference so you can click through and check.
It is off until an admin turns it on, in Team settings → Features → Answers from your work, because it is the one thing that lets the AI read task content. When it is on:
- It reads only what you can open yourself. The AI asks for lookups — a search, a task, a project's summary — and sjTasks runs them under your own access. A guest asking about a project gets a guest's view of that project and nothing about any other; a field worker's question is answered from their own jobs only. Nobody learns anything through the box that they could not read on a page.
- It reads a little, not everything. At most a few dozen tasks and the latest notes on each, never the whole team. The answer says what it read: Read from 12 tasks in Acme Redesign.
- It writes nothing. No task changes, no note is added, no date moves. If it cannot find an answer in what it read, it says so rather than inventing one.
- It costs more of the allowance than a filter question does — it has to read the tasks before it can answer — and one question is capped so it can never use a month's allowance on its own. Its answers are kept for ten minutes, because the work changes.
Turning it off
The box is on from the start, on every plan. An admin can turn it off in Team settings → Features → Ask in English; while it is off there is no box at all — nothing greyed out, nothing to wonder about — and the menus answer the same questions.
Each team has a monthly AI allowance that follows its plan — measured by how much work each question takes rather than by counting questions, so a short question costs less than a long one, and a question the team has asked before costs nothing. If a team uses its month up, the box says so politely and the menus keep working exactly as before — asking in English is a convenience on top of the filters, never a gate in front of them.
A question your team has asked before is answered from memory for a month, instantly and without using any of the allowance — Monday's "what did we finish last week?" only ever costs once.