Questions about starting off with document understanding

Hello all,

Please i have a few questions regarding DU

  1. what is the difference between using the orchestrator Document understanding module and the studio desktop document understanding package activities..?

  2. does ai center depreciating some out of the box ML packages mean I have to create my models each time? or ar DU models and AI models independant?

  3. if i have classic invoices, is it enough to use proposed out of the box models? or is it always preferable to create my models

  4. does document understanding consume anything?

There is classic DU and modern DU now.

Here is quick difference between both.

Classic

  • Component‑based, manual pipeline
  • Requires creating each step separately:
    Digitize → OCR → Classify → Extract → Validate → Export
  • Complex project structure

Modern

  • Unified, guided experience in Studio Modern
  • Document Understanding Process” is auto‑generated
  • Simplified activities with fewer manual configurations
  • Better for new developers

Yes, all out of the box AI models are deprecated except Document Understanding. For DU project you will be creating and consuming the model from AI Center for the remaining deprecated models, there is an alternative, GenAI activities - Activities - About the UiPath GenAI Activities package

I have recently spoke about it recently here:

Yes, both are independent.

It’s better to use predefined specialized model for invoices pretrained by UiPath. You will have to train less. If you want you can create your own model as well.

Yes, Document Understanding consumes AI Units whenever you use any AI/ML-based models.

Basic DU steps like OCR, Regex extraction, validation, and keyword classification do not consume AI Units.

very helpful thank you!!

so if I understand correct:

  • the classic method is using the document understanding packages on studio and building the flow ourselves
  • Or using the existing document understanding process template
    is it correct?

also i saw there is a document understanding module in orchestrator with both “classic” and “modern” approaches, and saw it can generate a workflow directly

what is the difference between the latter methods and the studio ones, or are we obligated to use orchestrator module?

edit: i started watching your video, when you talk about genai, are you taking in account document understanding features?

Not really!

In both classic and modern we use document understanding process template. The difference is in the using classifier and extractors.

In classic DU, if we want to use AI Center model we use Machine Learning Extractor . In modern DU we use Document Understanding Project Extractor

In classic DU we create AI Center ML Package
In modern DU we use a guided model-building.

We have guided model building in modern DU and if we choose classic, we will be able to see the project from AI Center for more control on each feature.

No, it’s a kind of replacement for the AI Center out of the box ML models we had which are now deprecated.