Dashboards and charts

AI chooses the chart. Your data supplies the numbers.

Describe the chart you need. Geist calculates its values from your existing event logs, imported files and connected business data. Wisp, the AI team in Geist, configures the view and explains the results using the chart's evidence.

Ask for the dashboard you need, in your own words.Wisp picks a template and binds it to your event log.Every tile reads the same event log.Wisp answers only from the histogram's own evidence.
From raw data

From an SAP log to a chart

Thousands of SAP change documents become one readable chart, automatically. Geist turns transaction codes into activity names, converts SAP dates and local times to UTC, sets aside rows it cannot use and counts them. Then it calculates each case's cycle time and draws the result.

  • Set up in minutes, not in a weeks-long project: Geist detects the columns, and you can override any mapping
  • SAP event logs over OData (beta), or any log as CSV or XES
  • Unknown transaction codes go to a review list instead of being guessed
  • Charts refresh after new imports, with five-minute checks as a fallback
How process mining works
Change documents arrive as SAP stores them: transaction codes and local dates.Each code becomes an activity, each date and time a UTC timestamp. A row without a valid date is set aside.Events group into cases, and each case gets a cycle time.Every case cycle time goes into the histogram, with median and P95.Key insights are calculated from the same numbers, not written by a model.
Data care

What happens to messy data

Geist flags problems in an event log and counts what it skips.

Missing fields

CSV rows without a case or an activity are skipped and counted, with the reason, and the import report shows how columns were mapped. Invalid timestamps stop a file import. Data-quality checks cover the fields available in the stored log.

Duplicates

A change document that SAP sends twice is stored once, keyed by its change number. Identical rows in a CSV or XES file stay separate events, so remove repeats from a file export before you import it.

Dates and time zones

Dates in common formats, including SAP's date and time fields, are converted to UTC. You set the time zone of your SAP system, and Geist records it with each event. When a file's timestamp needs converting, its original value stays with the event.

Outliers

Unusually long cases are marked, never removed from the numbers. If a few extremes flatten a chart, its axis can be cut off. Those marks stay visible and show their true value. When Wisp cuts an axis, it says so in its answer.

Chart choice

The question decides the chart

Wisp picks from 22 chart forms and variations. A histogram shows one duration across all cases, a box plot shows spread, and a Pareto chart shows when a few categories carry most of the volume. The server checks each preview. Wisp tries up to twice to fix what the checks flag and tells you about any warnings that remain.

  • Forms from line and bar to box plot, Pareto, funnel, waterfall and calendar heatmap
  • Preview checks for skewed values, long or crowded labels and too many series
  • Values come from the server, whichever form Wisp picks
Ask a question about the data.Wisp picks the form that fits the data and the question.The server checks each preview. Wisp tries to fix flagged issues and reports remaining warnings.Each question gets its own form.
Edit and export

Change it in words, export it when it is right

Ask Wisp to split a chart by another field, switch the chart type or move a tile. It changes the chart definition and shows a preview for review. A finished dashboard goes out as a 16:9 PowerPoint deck or a PDF, and each chart as Excel, CSV or PNG.

  • Visual check (beta): when available, Wisp reviews the rendered chart, tries up to two repairs and reports remaining warnings
  • Speaker notes carry the dashboard summary, and each chart keeps its stated limitations
  • A tile that failed to load is left out of the deck instead of filled with placeholder values
Ask for the change in your own words.Wisp changes the chart definition. The server computes every number from the log.Wisp looks at the rendered chart before you see it.Export the dashboard once it is right.Each slide names its source and when the data is from.
How it works

From a question to a deck in four steps

  1. Step 1

    Ask

    Say what you want to measure and how to split it, in English or German.

  2. Step 2

    Build

    Wisp picks a chart from the catalog and writes its definition. The server computes the values from the data source, with no generated SQL or code.

  3. Step 3

    Check

    Wisp previews the chart and uses visual checks when available (beta). It tries to fix layout problems and reports any remaining warnings. It saves the chart when you ask it to.

  4. Step 4

    Share

    Share the dashboard in your organization, present it, or export it to PowerPoint or PDF. Export individual charts to Excel.

FAQ

Questions about dashboards and charts

Can the AI make up numbers?

For governed charts, Wisp configures the chart definition and the server calculates plotted values from the selected, authorized dataset. The language model does not supply the numbers or write SQL or code. Data quality, filters and AI interpretations still need review before a decision. Forecasts are estimates, and sandbox-generated outputs follow a separate workflow.

What does AI analysis of a chart look at?

The chart's structured data: the first 200 rows, with up to 24 fields each. When that leaves more than 80 rows, the model reads the first and last 40 of them. Every note cites the chart, series or data source it relies on, and a note without a valid citation is dropped. Each note carries a confidence score. The analysis is stored with the chart revision and a fingerprint of its data, and is marked stale when either changes.

How current is a dashboard?

When an event-log import or Globe data change is saved in Geist, affected visible tiles refresh automatically. They also check every five minutes, when you return to the tab and after reconnecting. Refresh data reloads them on demand. Each tile shows the time reported by its source. ERP data first has to arrive through its connector or import schedule. Charts bound to a specific log keep that selection; historical chart snapshots and sandbox outputs stay unchanged.

What is in the PowerPoint export?

A 16:9 deck of the dashboard's charts, with the dashboard summary in the speaker notes and each chart's limitations. A tile that failed to load is left out. PDF comes from the print view, and each chart also exports as Excel, CSV or PNG.

Who can build charts, and who sees the data?

Anyone allowed to edit a dashboard can ask Wisp for a chart, without training. Sharing a dashboard does not widen access to the data behind its charts.

Next step

Bring a report you build by hand

Send us a recurring report and the data behind it. We rebuild it as a dashboard with you.