Skip to main content
Help Center

Analyse reports with AI

Drawing the key insights out of a report – via the Analyse with AI entry in the report's three-dots menu.

Prerequisites

  • A connected AI in the tenant (see Connect an AI)
  • A report you want to analyse

A report delivers figures. The question of what those figures mean, it doesn’t answer – you ask yourself that, usually in your head, often with the Excel export as a detour. The AI analysis shortens this: it reads the report and names what stands out.

How to do it

  1. Open the report as usual and set it up the way you need it – period, filter, grouping.
  2. Click the three-dots menu at the top right.
  3. Choose Analyse with AI – the entry sits directly below Export (report).

The result appears immediately in the AI analysis dialog above the report. No export, no second tool, no switching.

The Invoicing analysis report with the three-dots menu open at the top right; it contains the entries Enable multiple selection, Print, Export (report) and, highlighted below, Analyse with AI
The report's open three-dots menu with the "Analyse with AI" entry below "Export (report)"
The AI analysis dialog with the result: a brief assessment of the Invoicing analysis, structured into Overall picture, Anomalies as a list, Trend and Questions for further review; at the bottom the Close button
The "AI analysis" dialog with the result – structured into Overall picture, Anomalies, Trend and Questions for further review

The analysis thinks along with the report

This is the point that surprises on first use: The analysis is not the same feature with different numbers. It adapts to what kind of report it has in front of it – and delivers completely different insights for an hours statement than for an Invoicing analysis.

ReportWhat the analysis thinks about
Hours statementWho booked how much, where it runs out of hand, which activities eat up the time, where plan diverges from actual.
Invoicing analysisHow revenue is distributed, which customers carry it, where something drops away, which line items stand out.
Project analysisWhere progress stands against the budget, which projects are tipping over, where the contribution margin is thin.

That’s why it’s worth trying the analysis on different reports instead of testing it once on one and then forgetting it. And that’s why it’s worth setting the report up properly first: the analysis sees exactly the data you currently have displayed – no more. A tight filter gives a narrower but sharper answer.

What you do with it

  • Before the monthly review. Know the notable points before anyone asks about them.
  • As a second pair of eyes. The AI spots the thing that stands out – including the one you wouldn’t have looked for.
  • As a starting point. The analysis tells you where it’s worth digging. The digging then happens in the report.
  • Instead of the Excel detour. For a “quick look to see if anything stands out”, you no longer need a spreadsheet.

The limits

The analysis is an interpretation, not a calculation. It may emphasise an anomaly that isn’t one, or miss one. What it says is a pointer – the figures in the report are the truth.

It reads only what is there. It knows neither your back story nor the reason January is always weak. If a figure looks the way it does because of an arrangement from 2019, it sees an anomaly.

It phrases things anew each time. The same analysis twice gives two similar but not word-for-word answers. For figures that have to be exact and repeatable, the report is what counts – see Why teamspace relies on rules for AI.

Common questions & needs

You want to …How to
A sharper analysisFilter the report more tightly beforehand. The analysis only sees what is displayed.
Different insightsTake a different report – the analysis adapts to its kind.
To pass the analysis onCopy the result out. It is text; the report remains the reliable source.
To know what it is based onOn the report data currently displayed – limited by the maximum input length (see Connect an AI).
A recurring analysisA report is the right tool for that, not the analysis. It is for a look, not for routine.
To ask freely worded questions of the dataThe analysis can’t do that – for that there is MCP.

Notes

  • Large reports can be truncated. The maximum input length limits what goes to the AI. With very extensive reports the analysis may not see everything – filtering more tightly helps.
  • Only the displayed data goes out, and only when you call the analysis – see Your AI, your data.
  • The analysis changes nothing. It reads but does not write back.
  • The report remains authoritative. If the analysis and the figures contradict each other, the figures apply.