Box unlocks agentic data extraction to help companies dig up insights faster

Enterprise content management provider Box Inc. is expanding its armory of artificial intelligence tools with the launch of Box Extract in general availability.

The new tool uses third-party AI models from companies including OpenAI Group PBC, Google LLC and Anthropic PBC to extract valuable insights embedded in documents such as invoices and contracts to enhance enterprise decision-making.

Box believes that the new tool is going to become vital for enterprise customers. It’s well=known that there’s vast amounts of organizational knowledge locked inside documents such as spreadsheets, product listings, product specifications, policy documents, videos, charts, social media and other types of content. Yet most organizations are unable to access the insights held within this data because they rely on manual processes and legacy tools that simply cannot scale up.

AI agents can make a difference, Box says. Box Extract, which was announced in September, provides access to two kinds of data extraction agent, including the Standard Extract Agent that’s focused on rapid, cost-efficient data capture, and the Enhanced Extract Agent that has the deeper reasoning capabilities needed to conduct more exhaustive searches for information.

The two agents, which can be powered by the customer’s choice of large language model, including Google’s Gemini 3, OpenAI’s GPT-5.2 and Anthropic’s Claude Opus 4.5, provide enterprises with access to a robotic worker that’s able to dive into their content far faster than any human could do. The agents are designed to leverage the models’ reasoning skills to understand the documents they’re looking through, and pick out the most vital information within them to inform enterprise operations.

Box co-founder and Chief Executive Aaron Levie said many enterprises are sitting on a gold mine of insights that live within their untapped content, but only a few know how to access them. “With Box Extract, that information is now unlocked and can transform how businesses analyze information and make decisions,” he promised. “By turning unstructured content into structured, usable data, organizations can deliver real-world impact by having their content actively work for them across their most important lines of business.”

Box Extract is available through the company’s Enterprise Advanced subscription plan and is said to excel at understanding document structure and meaning. By breaking down each document it scans into components such as paragraphs, tables and charts, it can pull out the most relevant information that organizations need to know.

Besides the two standard extraction agents, customers will also be able to create customized Extract Agents that are better-suited to their organizational needs, the company said. Box said whatever insights they dig up will be stored in a separate destination within Box, where organizations can access it to make faster decisions, automate workflows in Box, streamline content discovery and more.

The update promises to be one of the most powerful new AI features Box has introduced so far, and it’s also long overdue, said Holger Mueller of Constellation Research. For all of the advances in AI so far, the reality is that the onus is still on humans to understand and organize enterprise documents. “Enterprise workers often find themselves challenged even to find the right document, which could be located anywhere inside a mass of different folders,” the analyst explained. “Box Extract promises to put an end to these hassles, enabling business users to find the right piece of data or the intelligence they require to inform their decision making in seconds via a conversational experience. This will accelerate how people work with documents, and that’s a significant upgrade for the future of work.”

Box highlighted a number of example use cases for Box Extract. For instance, it said financial services providers could use the service to enhance loan originations and servicing by extracting details such as loan terms and due dates. Government agencies could use the tool to extract details such as permit types, fees and inspection dates from public records, grants and contracts to accelerate compliance and service delivery.

Insurance providers are another target market, with Box Extract able to extract critical information from accident reports, accident images, hospital bills and so on to aid in claims processing. Legal teams can also make use of Box Extract by processing long contracts to identify risky language and clauses that they might need to be aware of.

Image: Box