For AI agents: the complete documentation index is available at https://docs.clickmax.io/en/llms.txt, the full documentation bundle is available at https://docs.clickmax.io/en/llms-full.txt, and this page is available as Markdown at https://docs.clickmax.io/en/features/max/build-insights-dashboard-with-ai.md.
  • English

      • How do I ask the AI to build an Insights dashboard?

      Describe the numbers you want to track and the AI builds the dashboard under Commercial → Insights, with the widgets grouped in sections. The dashboard arrives saved, and you keep editing it on screen: change a widget's chart, resize a card, change the period.

      Before you start:

      • Role: Owner, Administrator, Editor, Viewer, Guest, or Attendant. Finance-only roles do not see Insights.
      • Know what you want to read on the board: which indicators, from which pipeline, over which period.
      • To use your own AI instead of Max, see the MCP reference (Portuguese).

      1. Say what the board has to answer

      Ask for the numbers, not the charts: "I want to track won revenue, opportunities created, and the funnel by stage of the Sales pipeline". The AI picks widgets from a closed catalog of indicators — it does not invent a metric. If an indicator you asked for is not in the catalog, it says so and offers the closest one.

      Worth asking for as well:

      • A ready-made template — "a board to follow the sales team" delivers one of the templates instead of a custom build.
      • Pipeline — indicators broken down by stage (funnel by stage, opportunities by stage, open value, time per stage, loss reasons, MQL → SQL) belong to a single pipeline. Name it: without it the AI asks instead of choosing for you, because a guessed pipeline makes the card show one pipeline's number while you read it as the whole operation's. When you ask for a ready-made template, the pipeline you name (or the most recent one, when you don't) is stamped into each card — the board never switches pipeline on its own when you create a new pipeline later.
      • Period — "current month", "last 7 days", or specific dates. The shortcut is saved as a shortcut: when the AI reads the board's numbers later, it reads the shortcut's window recalculated right then — the same one the screen shows — not the whole history.
      • Goals — asking for the comparison against the goal turns on the dashboard's Goals toggle, and each card starts showing its pace against the registered goal.

      2. The board arrives grouped in sections

      The AI builds the dashboard in one go: it creates the widgets, groups them into sections with titles ("This month's result", "Team and reasons"), and sets each card's width and height so the rows close — the same way a ready-made template does. That is why asking for the whole board in one sentence beats asking widget by widget.

      Dashboard limits: 40 widgets per dashboard (each section title takes one of those 40 slots) and 25 dashboards per account. When you hit the cap, the AI tells you and you choose what to remove.

      The filter at the top has caps too: up to 100 items per list (pipelines, stages, owners, sources) and up to 200 conditions in the advanced filter. These are safety ceilings, far above any normal selection — if you hit one, filter by the items that matter instead of ticking everything.

      3. Check the numbers before trusting the board

      Open the dashboard and review:

      • The indicator is the one you asked for — the card title says what it measures; check the widget's gear if the reading looks off.
      • Cards showing "Could not calculate" — that widget's calculation failed. Ask the AI for another indicator, or fix it in the gear. Asking for an indicator that does not exist never becomes a card: the AI is told right away and offers the alternative before saving.
      • Cards at zero — zero is an answer too: it may be a cut with no results in the period. Compare against the filter at the top before concluding the data is gone.
      • Ignored-filter notices — some indicators do not use every filter from the top (overdue activities are a snapshot of now, not of a period). The card states which filters it ignores.
      • A cut that is not on the filter bar — the AI cannot save a cut the screen has no control for (status, temperature, priority, value, origin, task type): those go into the advanced filter, which stays visible and you can clear. On boards created before that, such a cut shows as an amber badge on the filter bar (for example Status: won) with an ✕ to remove it — if the numbers looked low for no reason, that was it.
      • Widgets waiting for a parameter — breakdowns by custom field or by tag need the field or up to 8 tags. When asking the AI, name the field or the tags: without them it cannot create the card and will ask. Building on screen, you pick it in the widget's gear.

      4. Adjust the board afterwards

      You can keep going by chat or on screen:

      • Change a widget's chart (Bar, Line, Pie, Ranking, List, Number, Table).
      • Resize the card — width and height, on the card handles.
      • Add a title to open a new section.
      • Reorder — the card order is the order shown on screen.
      • Change the period, Goals, shared filter, and Live mode without touching the widgets.

      The filter belongs to the dashboard, not to the widget: a period or a pipeline chosen at the top applies to every card at once. Cuts such as open/won/lost are a choice of the indicator itself, not of the filter.

      If it did not work

      • You cannot find the dashboard that was created: it lives under Commercial → Insights, in the dashboard list on the left.
      • The AI said the indicator does not exist: the catalog is closed, so ask for the number from another angle ("revenue by owner" instead of "revenue by city"), or build the reading with the filters at the top.
      • The AI said it needs the field or the tags: breakdowns by custom field and by tag only exist with that parameter — answer which field or which tags and it finishes the job.
      • A card shows "Could not calculate": that metric combination does not run — ask the AI for the alternative instead of reading the zero as a result.
      • The numbers do not match what you expected: check the period at the top, whether there is an amber cut badge on the filter bar, and whether the card is warning that it ignores a filter.
      • A stage card is talking about another pipeline: open the widget's gear and pick the pipeline. Every stage card has its own — two cards on the same board can look at different pipelines.
      • You hit the widget or dashboard cap: remove what you no longer use and ask again.
      • An action was blocked by an error: see the error codes reference (Portuguese).