# What you can do > Self-service analytics on a serious engine: connect what the company already runs on, get your own answers without filing a request, and put the dashboard where the work happens Source: https://aipril.dev/docs/analytics/what-you-can-do ## What it is Analytics people can use themselves. Connect the systems the company already runs on, look through the data, chart what you found, and assemble the dashboard the team will open every Monday - all of it without writing a query, filing a request, or waiting for somebody with the right access to come back from holiday. Underneath is a modern analytical engine: it works in memory, on very large volumes, and it counts where the data lives, not in your browser. That is the difference between a report that opens on tens of millions of rows and one that opens a spinner, and the reason nothing has to be cut down to a sample first. ## AI-first, and still yours to manage Analytics tools used to begin with learning the tool: which of forty settings to tick, in what order, and what the one with the ambiguous name actually does. That part is over. Say what you want and the assistant does the setting-up, while you look at the result instead of at a panel. Nothing is locked away, though. Everything the assistant touched is an ordinary setting you can open and change - the sorting, the number format, the filter, the colour rule. And you do not have to know a formula language: where a formula is wanted, smart suggestions offer it in terms you can check. If you are the analyst who knows which expression you want, write it: the full depth is there, and now it is optional. > AI-first, people-managed. The assistant gets you to a working screen in one sentence, and every > choice it made along the way is one you can see, change or overrule. What it will do for you: a chart from a prompt, a whole dashboard assembled in dialogue, a formula written on request, a rule applied down a column, the values that arrived in three spellings tidied into one. It works on real data, inside your own access rights, and shows you the result before anything is saved. The whole of it is on its own page: [AI in analytics](https://aipril.dev/docs/analytics/ai.md). ## Connect data that already exists - **Any database** - PostgreSQL, Oracle, Microsoft SQL Server, MySQL, ClickHouse, Snowflake, MongoDB and others. The list is not a closed one, and connecting one is a setting rather than a development project - **Any service with an API** - if a system can be asked for its data, it can become a dataset, and analytics over it does not wait for a data pipeline to be built first - **Files people actually work in** - upload a spreadsheet, or have files collected automatically from a folder, a file server or cloud storage, so the numbers somebody maintains by hand stop being a separate world - **A built-in data lake when you need one** - data can be loaded into the platform's own storage instead of queried in place, for sources that are slow, fragile, or under too much load to be asked analytical questions. Same model either way: moving a report between the two is a decision about load, not a rebuild - **Freshness you set once** - how often each source is collected and its figures recalculated is a setting per dataset, from minutes to overnight, with a refresh-now button for the times it matters. Every dataset says when it was last updated, so "is this current?" is answered on the screen instead of by asking around - **Your own data, live** - requests, tasks and statuses from your processes are read where they stand, with no schedule and no sync window between something happening and the report showing it ## Find the answer yourself The exploration table is where the day is spent, and it is built to end the day inside it: - Sort, filter, reorder and resize; save the setup you like as a preset and come back to it - Colour, icons and data bars applied by rule, so what is out of range finds you - Your own columns from a formula - a share, a difference against last period, days overdue - without asking anyone to add them to the report - Mini charts inside the cells, pivot, and drill-down into what a number is made of - Export to Excel when somebody insists on Excel - Works on a phone ## Build the dashboard - Drag the widgets where you want them, resize them, copy them between dashboards - **Cross-filtering** - click a bar and the whole screen follows: the other widgets re-filter around what you just picked, with no page reload and no second report - Filters over the screen and drill-down inside each widget, so "and by channel?" is answered in the meeting, not after it - More than charts on the grid: tables, KPI tiles, a calendar, a schedule, a Gantt, a map, a live task board from your processes - Send a link and the person opens the same live screen, not a picture of it ## Put it where the work happens > A BI tool answers the question. Being part of the platform decides where the answer lands - on > the page where the request is approved, next to the record it is about, in front of people who > would never have opened an analytics tool to look for it. - Drop a chart onto a page of an app or an internal portal - same widget, same numbers, no export - Stand it beside the thing it is about: the approval, the request, the client card - Put it in front of people who do not use BI at all: a manager sees the figure on the portal page they already open every morning - Hand the same data to an external AI assistant over [MCP](https://aipril.dev/docs/platform/integrations/mcp.md), under the same access rules ## From the number to the next action A BI tool that stands on its own ends at the chart: somebody spots the deviation and then leaves - into a messenger to ask about it, into a task tracker to assign it, into a third system to do something. Here the finding can be discussed, assigned and acted on without leaving the screen it turned up on, because the CRM, the request queue and the project portfolio are apps on this same platform, not systems the analytics is pointed at. That is the difference that does not fit in a bullet: [Inside the work](https://aipril.dev/docs/analytics/inside-the-work.md). ## Every chart you need, and the one you invent - Bar, column and combo; line and area; pie and donut; KPI cards; waterfall; scatter; radar; treemap; heatmap; range; map - Small multiples for bar, pie and radar - one shape repeated per category, instead of five series fighting over one axis - Shared presentation settings, so a dashboard reads as one screen instead of a pile of unrelated reports - **Your own chart type** - a shape the catalogue is missing is a simple definition, not a feature request. Write it and it renders against your data alongside every built-in type, or describe the chart you want and have the AI write the definition for you ## One set of numbers for the company - **Revenue means the same thing on every screen.** How a metric is calculated is stated once and read from there by the chart, the table, the Excel export and the AI assistant. Arguments move from "whose number is right" to "is that the right definition", and that one gets settled - **One dashboard, different rows.** A regional director does not need a copy of their own: the access rule sits on the data, so everyone opens the same screen and sees their own part of it - **Columns not everyone may read** can be hidden without building a second version of the report - **The classic silent error is caught, not served.** Put sales and stock on one screen and a naive join repeats every sale once per stock snapshot, inflating revenue with no warning. The platform recognises that shape and counts each side at its own grain; where a figure genuinely cannot be computed across two facts, it says so instead of returning a plausible one. Details in [Under the hood](https://aipril.dev/docs/analytics/under-the-hood.md) ## Where it is used | Situation | How it is answered here | |---|---| | The weekly review opens with an argument about whose revenue figure is right | The metric is defined once and every screen reads that definition. The argument moves to the definition and is settled once | | Every regional director wants their own copy of the dashboard, filtered to their region | One dashboard. The access rule sits on the data, so each of them opens the same screen and sees their own region | | The report is missing one column - days overdue, share of total, change against last period | Add it yourself in the table. It needs nobody's approval and can go straight onto a chart | | Data sits in a database, a spreadsheet somebody maintains, and a system that only has an API | All three become datasets and join into one picture - no pipeline project standing in front of the first answer | | Operations wants open requests and the money behind them on one screen | The task board is a widget like any other, so live process records and figures sit side by side | | A manager will not open a BI tool, but does open the portal every morning | The chart goes on the portal page, and the report reaches the person the report was for | ## Next [Under the hood](https://aipril.dev/docs/analytics/under-the-hood.md) - what the engine actually is, and why the number in the chart matches the number in the export.