Americas

  • United States

Asia

mfinnegan
Senior Reporter

Adobe’s new Acrobat genAI assistant aims to streamline reading PDFs

news
Feb 20, 20243 mins
Artificial IntelligenceGenerative AIProductivity Software

The generative AI assistant provides content summaries for PDF and other file types, with users able to pose queries relating to a document in natural language.

Acrobat AI
Credit: Adobe

Adobe is promising to make it easier to digest lengthy PDFs with a new generative AI (genAI) assistant coming to Acrobat.

The Acrobat AI Assistant is now available in public beta for customers on paid Acrobat individual and teams licenses (a private beta is in place for enterprise customers), as well as the free-to-use Reader app. There’s no immediate word on cost at launch, though Adobe said an add-on subscription will be required for Reader and Acrobat users.

“The AI Assistant in Acrobat is a helpful feature for existing teams using Acrobat and for potential new clients working with and managing information in lots of text-heavy documents  (legal, research, editorial, archival institutions etc.),” said Rosanna Jimenez, research analyst at S&P Global Market Intelligence.

“While there are open-source applications that can be useful for one-off use cases that do much of the same, having a standardized AI tool for teams/organizations will enable more cohesion and accuracy,” she said.

Accessed via a sidebar in the app, the Acrobat AI Assistant provides an overview of a document with short summaries of content in various sections. In addition to information on text content, the summaries can also provide details from graphs and infographics. By clicking on a summary, users are brought to the relevant section for further detail.

Users can also ask the AI assistant questions relating to the document from the sidebar, with suggestions provided. The resulting answers contain citations with links to the information in the document.

There are limits to the answers the AI assistant will offer, as it is restricted to responding based purely on information from a document, rather than providing additional context sourced from the web.

Another feature is the ability to format summarized information so that it’s ready to be shared via email or other means.

Adobe stated that no documents analyzed by the AI Assistant are stored or used for training its language models.

These are just the initial features for the Acrobat AI Assistant. Adobe also will enable the Acrobat to summarize a wider variety of document types as well, including Microsoft Word and PowerPoint files.

There are plans to target the AI assistant at authoring and editing within Acrobat, with the ability to generate first drafts, for example, or to change voice and tone.

Adobe will also integrate Acrobat with its Firefly generative AI tool to add images to text documents.

Ritu Jyoti, IDC’s group vice president for Worldwide Artificial Intelligence and Automation services, expects strong demand for the Acrobat AI Assistant, with knowledge discovery one of the “top strategic use case categories for genAI.”

While hallucinations are a concern (as with any genAI service), Jyoti said Adobe minimizes them by grounding responses within the document content and providing verifiable results with attribution.