AI-assisted form prefill

What it is

Filling in a task form from a document is slow and error prone work: someone reads the file, finds the right fields, and types them across. Flowable Work now does the reading for you. Upload one or more documents to a task and AI-assisted analysis proposes values for the form fields, so the user reviews and confirms rather than transcribes.

The AI form assisstant shows the extracted values from the document(s) and the user can decide, which values they want to use in the form. Where a field has more than one plausible source in the document, the user can choose between candidate values rather than being given a single guess.

What’s included:

• AI-assisted document analysis that prefills task forms from uploaded documents.
• The AI assistant proposes changes to the form which the user can accept or reject.
• Support for choosing between multiple candidate values when a field has more than one source.
• Separate mapping for each content model, so different types of documents can fill different sections of the form.

Why it matters

The time between receiving a document and having usable data is where most manual effort and most data entry errors live. Letting AI propose the answers and asking a person only to confirm them collapses that gap, while the proposed, existing and confirmed states keep a human firmly in control of what is committed. The result is faster handling of documents with a human still in control.

How it works

This feature can be enabled on a particular form by enabling document analysis on an attachment component in the form. The modeller configures the document agent responsible for classification and data extraction and sets the mapping from the extracted data to the form fields. When a user then uploads documents to a task, Flowable Work sends them for AI-assisted analysis to the configured document agent and maps the returned values onto the form’s fields. The AI assistant proposes values that the user can accept or reject, field by field. Nothing is committed until the user confirms, so the prefill accelerates entry without removing oversight.

Example

A user opens a task that needs a dozen fields populated from multiple documents. Instead of reading and typing, they drag the document onto the task. AI analysis proposes values for most fields, flags two fields where it found more than one candidate, and leaves an already-populated field marked as existing. The user picks the correct candidate for the two ambiguous fields, confirms the rest, and completes the task in a fraction of the usual time.