Geist
Book a demo
Enterprise import & migration

Migrate the process library without redrawing it

Migrations stall when every diagram has to be redrawn by hand. Put ARIS exports, Visio files, PDF handbooks and Word descriptions into one session of up to 200 files, and Geist turns them into BPMN 2.0. Where the AI had to interpret a source, the model waits for a person to review it.

Book a migration demoRequest access
An ARIS export goes in. Visio, Signavio, draw.io and PDF files import too.Org units become lanes.Functions become tasks. The events between them fold into the flow.The XOR rule becomes an exclusive gateway with both branches.Layout places every shape in its lane.QA scores fidelity. Imports that need a person go to the review queue.
One session, every source

ARIS, Visio, PDF and Word in one batch

Drop the exports from every tool your teams have used. A session holds up to 200 files of up to 60 MiB each and up to 2,000 pages once multi-page files are expanded. Identical files are recognized by their content and imported once.

  • VSDX, AML from ARIS, PDF, DOCX, PPTX and XLSX, plus JPG, PNG, GIF and WEBP images read by a vision model
  • Single files through File > Import: BPMN, Signavio SGX, draw.io, XPDL, VDX, DOCX, AML and ZIP
  • Within a process, import enrichment points out activities that look like duplicates
  • Preview: merging the same process found in several files into one model
One session takes the whole batch, up to 200 files.ARIS and Visio files map their shapes and swimlanes to BPMN.An identical file is recognized by its content and imported once.PDF handbooks are interpreted by AI, so their models always go to review.A Word description with a gap gets a question before anything is generated.The run continues on its own. Review collects what a person has to check.
Text sources

Processes that exist only as documents

Many processes were written down in a Word file or a handbook and never drawn. Geist reads numbered paragraphs as candidate steps. Where the text leaves something open, the AI asks a short question before it generates the model and saves your answer with the source.

  • Pick a suggested answer or write your own
  • A document that describes no process is flagged instead of forced into a model
  • Every model drawn from text goes to review with the source beside it
Numbered paragraphs in the description become candidate steps.Paragraph 3 names two teams for one decision.Before generating, the AI asks instead of guessing.The answer is saved with the source file.The generated model goes to review with the source beside it.
Review inbox

Every import shows how faithful it is

Each file lands in the review inbox with its source page next to the generated model. Imports with a fidelity score below 0.86 wait for a person, and below 0.72 they also carry warnings. Open the model, check it with Wisp, the AI team in Geist, or run the import again.

  • Filter by Ready, Needs review, Warnings or Failed
  • Open every converted model in Finder at once, or download the run report
Each imported file lands in the review inbox next to its source.The status comes from a fidelity score that compares each model with its source.Below 0.86, an import waits for a person to review it.Below 0.72, the import also carries warnings.When the run ends, you open every model in Finder at once.
How it works

From a library export to reviewed models in four steps

  1. Upload

    Drop the library

    Drop a folder of VSDX, AML, PDF, Office or image files. Each file gets its own row with progress, cancel and retry.

  2. Clarify

    Answer what the sources leave open

    When a description is ambiguous, Geist asks a short question before it generates, instead of guessing.

  3. Review

    Check what the AI interpreted

    Structural QA checks every model. Models from PDFs, images and written descriptions always go to review, with the source beside the result.

  4. Results

    Open the models in Finder

    Converted models open in Finder with their source system recorded. Each run and its files stay listed under Import runs.

FAQ

Questions about large migrations

How large can one migration be?

One session holds up to 200 files of up to 60 MiB each, and up to 2,000 pages once multi-page files such as Visio documents are expanded. A larger library runs as several sessions, and every run stays listed under Import runs.

Is a PDF converted automatically?

A PDF or an image is interpreted by AI, so every model drawn from one goes to review with the source page beside it. ARIS and Visio files map their shapes to BPMN and get a fidelity score; below 0.86 they wait for review too.

What happens to duplicate files and processes?

Identical files are recognized by their content and imported once. Within a process, import enrichment points out activities that look like duplicates. Merging the same process from several files into one model is in preview and not yet part of a migration run.

Where can the files come from?

From a local upload, the files already in your workspace, or the Import Hub sources for Visio, ARIS and draw.io. SharePoint and Google Drive connectors bring files in as well; both are in beta.

Which ARIS objects does the import map?

Functions become tasks, events become BPMN events and rules become gateways. Org units, positions and roles become lanes, process interfaces become call activities and information carriers become data objects. Where an EPC is ambiguous, the import says so in its diagnostics.

Does the import learn from corrections?

Yes. Improve import records your corrections, such as an industry, an ARIS remap or a label fix, and later imports in your organization reuse them. Labels keep their source language; German and English labels are normalized and internal codes expanded.

Related

Where the imported models go next

Next step

Migrate a sample of your library

Send us a set of exports, your own or a client's, and we run them through the migration with you.