[Category]Automation
Assistant
Ask the built-in assistant in plain language, and let it do the work for you.
The assistant is the chat built into Rocketbatch. You ask a question in plain language — "what's on my plate today?", "who haven't I contacted in a while?" — and it answers from your own contacts, tasks, deals, mails and projects. Results come back as cards you can open with one click, and anything that would change a record is put to you first as a plain yes-or-no question.
Opening the assistant
The assistant is always one click away. It opens over whatever you are working in, as a tab named after your first question, and the module underneath stays exactly where it was.
How to use it
- Click the lightning button at the bottom of the menu bar — its tooltip says AI Assistant. The panel slides in over the module you are in, with the question box ready.
- Type your first question and press Enter. The conversation opens as its own tab, named after the question you asked, and the module you were working in stays open behind it.
- To begin again from scratch, click the plus in a circle in the header. The conversation is cleared, the starter questions come back, and what was said before is no longer taken into account.
- To have the assistant beside the application rather than on top of it, click the pop-out button in the header. The conversation moves into a window of its own, which you can move and resize. The arrow-into-a-corner button in that window's header puts it back as a tab.
Good to know: several conversations can run side by side as tabs. The window version also offers a dark colour scheme and a switch that keeps the details of every step unfolded. You can start a conversation from the browser's search box too — that is described in the Browser chapter.
Related features
Asking a question
There is nothing to learn: no commands, no field names. You write the way you would write to a colleague, in German or in English, and the assistant answers in the language you wrote in.
How to use it
- Type your question into the box at the bottom marked Ask anything… and press Enter. Shift+Enter makes a new line instead of sending.
- Your question appears on the right, a working line with a running timer appears underneath, and the answer follows on the left — with result cards below it when records were found.
- Instead of typing, you can click one of the question chips under the header. They are matched to the part of the application you are in — "Show my overdue tasks" in Tasks, "Which leads are hottest right now?" in Deals — and the question is sent straight away.
- To use something you already asked, put the cursor on the first line of the box and press the Up arrow; press it again to go further back, Down to come forward again. The recalled question can be edited before you send it.
- If an answer is taking longer than it is worth, click the square button at the right of the question box. Work stops, a short Stopped note is added, and you can ask the next question at once.
Good to know: the chips disappear as soon as the conversation has started — they are a starting point, not a menu. An edit you make to a recalled question survives walking further back and coming forward again, and walking all the way forward returns whatever you had half-typed. If nothing happens for thirty seconds, the attempt stops by itself with a note suggesting you try again or ask something narrower.
Related features
How the answers look, What the assistant can see
How the answers look
An answer is a short piece of text, and under it the records it is based on — one card each, with the title as a link. So an answer is never a dead end: whatever the assistant found, you can open.
How to use it
- Read the answer at the top of the block. Underneath it, each result is a card with a number, its title, a line or two of details, and the links Open, Use this and Copy.
- Click the title of a card, or the Open link under it, to go to the record itself. A contact opens as a contact, a mail in the mail window, a note in the note window, and any other activity in its side panel. The conversation stays as it is.
- Only five results are shown at a time. Click Load more under the list to add the next few; the heading counts up as you go.
- Once everything already fetched is on screen, the link reads Load more from the server and goes and gets the next batch. When there is nothing left, a line says so and the link disappears.
- You can also ask for more in words — a follow-up such as "and the next 5" continues the same list where it left off.
Good to know: if your follow-up names something new — a person, a subject — it is treated as a fresh question rather than as the next page. And if there is no list to continue, the assistant says so instead of inventing one.
Related features
Asking a question, What the assistant can see, Checking an answer
What the assistant can see
The assistant knows which part of the application you are in, which record you have open, and which rows you have ticked in a list. That is why "this contact" or "the selected tasks" needs no further explanation. It does not see the rest of your screen, and it only looks up what your question calls for.
How to use it
- Tick the rows you mean in the contact, task or deal list, then open the assistant. The header shows how many records are selected, for example 3 selected.
- Say what should happen to "the selected" ones. The assistant works on exactly those records instead of searching for them again.
- If nothing is ticked, the rows you have highlighted are used instead.
- To make one particular result the subject of the conversation, click Use this on its card. The assistant confirms that the record is now in context and asks how you would like to proceed; your next question is understood as being about it.
Good to know: if the assistant had asked you which record you meant, Use this answers that question instead — your original request simply carries on with the record you picked. When you have a web page open in the built-in browser, the assistant can read that page as well; that is described in the Browser chapter.
Related features
It asks before it changes anything, How the answers look
It asks before it changes anything
Reading your data the assistant simply does. Changing it, never on its own. It says what it intends to do — the record and the values spelled out — and waits for you.
How to use it
- Ask for the change in the normal way, for example "mark these as done".
- Instead of doing it, the assistant answers with what it would do and a block that says Waiting for your go-ahead, with two links underneath.
- Click Yes, do it — or simply write yes. The change is carried out and the assistant reports what it did, beginning with Done.
- Click No, or write no, and it answers that it will not do it. Nothing has happened.
Good to know: anything else you write is taken as a correction to the proposal rather than as an answer to it, so "actually, make it next Monday" changes the offer instead of accepting it. An offer you have not answered is still good when you come back to it later the same day.
Related features
What the assistant can see, Keeping a conversation for next time
Checking an answer
Every answer can be taken apart. The grey lines above it are the steps the assistant took on the way to it, and each one unfolds.
How to use it
- Click one of the grey step lines — Searching records, Listing activities and the like. It unfolds and shows exactly what was asked for and what came back. Click it again to fold it away.
- A tick at the left of a step line means that step has finished.
- Beside the answer itself, click thumbs up or thumbs down to say whether it was any good. The thumb you pick stays marked.
- A thumbs-down also notes the answer for review, and the assistant says so in the conversation.
- The clipboard icon next to the thumbs copies a report on how the answer was produced, ready to paste into a message.
Good to know: if an answer could not be recorded, the thumbs stay greyed out and the tooltip explains why — the report can still be copied. In the pop-out window, the eye button keeps the details of every step unfolded from the start.
Related features
Keeping a conversation for next time
Something the assistant worked out once can be kept, so you do not have to ask for it again.
How to use it
- With the run finished, click the bookmark button in the assistant header.
- Give it a Name and a short Description of what it does, then click Save.
- The steps the assistant took are saved as a workflow for your workspace, with the same values it used. The conversation confirms it was saved.
- From then on it sits with your saved workflows, and the assistant can run it from your next question.
Good to know: if the assistant has not actually done anything yet, no dialog opens — the conversation simply says there is nothing to save. Saved workflows themselves are managed in the Workflows chapter.