Tables, charts, the work panel and replay

​

The agent draws, it does not retype

When you ask the agent something about your numbers, it reads them and then draws them: a headline figure, a table of the rows behind it, and a trend line where there is one. Its written reply sits above that and says what the numbers mean — what moved, what it implies, what to do next. This split is deliberate. A model asked to restate a forty-row table in prose does it at length, and occasionally retypes a number wrong. Handing it the rows and drawing them separately means the figures you read came out of the measurement, not out of a sentence.

​

What it can draw

The agent does not write its own markup. It names a component from a fixed set, and the app draws it: Two things follow from this. The agent can only produce layouts that have been designed, so a generated answer is held to the same visual standard as the rest of the app. And an app that has not shipped a component yet falls back to a plain-text description of what it would have said, rather than breaking the conversation.

​

Gaps are gaps, not zeros

A day nobody sampled shows as a break in the line, never as a drop to the floor. This is the same rule the dashboard follows. A share of voice of null means the prompt was not read that day; a share of voice of 0 means it was read and you were not mentioned. Those are different facts, and drawing the first as the second would invent a collapse that did not happen. The caption under a chart says how many days in the window were not measured. Movements follow the same rule. A change inside the sampling noise is reported as “no change” rather than as a small number with an arrow.

​

Copying and downloading

Every table and chart the agent draws carries Copy and Download.

• Copy puts the rows on your clipboard tab-separated, which pastes straight into Sheets or Excel and still reads correctly if you paste it into a message instead.
• Download saves a .csv. Charts save the rows behind the line, not a picture of it — a picture is not something you can pivot.

Only the table or chart is copied. The tool calls above it and the agent’s reasoning are not part of what you are taking away.

​

Pin to board

A chart or table drawn from your own data also carries Pin to board. Pick one of your boards, or let Attensira make one called My board if you have none. The pin lands on that board under Analytics → Boards. What is pinned is the read that drew the view, never the numbers or a picture of them. The board reads it again each time it opens, so next month it shows next month’s numbers.

• Only views drawn from a read of your data can be pinned. A draft, a card, or anything drawn by a tool that fetched the web or spent credits has nothing that could be read again, so it has no Pin to board.
• Only your own boards take pins. The boards Attensira builds in are not offered.
• A board holds up to 24 pins. A full board is listed but cannot be picked.
• If a pin fails, a message says so and the card stays as it was.

​

Seeing what it did

While the agent works, each step appears as it happens — reading your data, searching the web, fetching a page, changing what you track — each with its own mark, so a lookup is distinguishable from a crawl at a glance. Once the steps have finished they collapse into a single line, and the answer sits directly under the question. Click that line to open them again. Two exceptions never collapse: a step that failed, and a step waiting on your approval.

​

Worked for

When a run ends, one quiet line under it says how long it worked: Worked for 2m 14s. It counts from the run’s first step to its end, less any time it spent waiting for your answer to a question, and shows at most two units, the larger first. A run that took less than a second says Worked for under a second.

​

The work panel

The work panel holds what a run made, beside the conversation about it. Open it with Work at the top of the chat, and close it with Hide work; Replay opens it on the Replay tab. On a window at least 1,024 pixels wide the panel sits beside the chat, and opens on its own when the run made something to take away. On a narrower window it opens over the chat, only when you ask for it. A tab appears only when it has something in it, so a run that only answered a question usually shows Replay alone. Nothing is published from the panel. Review and publish goes to the work item’s approval in the Inbox, where your approval mode applies as it does everywhere else.

​

Watching a run back

Any finished or running piece of agent work can be watched back step by step. The replay is the same wherever you open it:

• Replay on a line in a project’s feed, including a Plan updated line;
• Replay on a run in an automation’s history, and on a test run once it starts;
• a deliverable, and an item in the Inbox, open the run that made them;
• in a chat, the Replay tab beside the conversation.

Each of these opens exactly the run it belongs to, never a neighbouring one. The header says who did the work (the crew role, in a project), which project it was for, when it ran, and its Because line: why it started, in one sentence built from the schedule or the change we recorded, never from a model’s summary. Under it are the run’s totals, how long it was working, the credits it spent and how many steps it took, and then the steps in the order they started. Each step says what was done and to what (“audited example.com/pricing”, “asked for your approval”), how long it took, the credits it was charged (free reads show none), and how it ended. A step that failed shows the reason in plain words. A stop for you shows how it was settled: approved, declined, answered, withdrawn, never answered, or still waiting on you. Beside a step is what it saw or made: Runs do not take screenshots, so the screenshot slot of a page says No screenshot for this step rather than showing an empty box. In the dialog, play steps through on their own, drag the scrubber, or use the keyboard: the arrow keys move a step, Home and End jump to the first and last, and Space plays or pauses. A replay carries at most 500 steps. A longer run shows the first 500 and says so; its totals still count every step. The replay is a summary of the run, not its raw log: a step shows its one-line description rather than everything the tool was called with, and a failure shows the plain-words reason rather than the internal error.