Box announces upgrade to Box AI, integration with GPT-4o

Box has unveiled a new set of features in Box AI that includes an integration with GPT-4o, support for image and spreadsheet files and the Box AI for Metadata API.

In addition, the cloud content-management company announced that end user queries in Box AI for Hubs, Documents, and Notes will be unlimited for organizations that are part of its Enterprise Plus plan.

Box said the following enhancements to Box AI, a suite of capabilities launched last year that comprise of generative AI models natively integrated into the company’s Content Cloud, will take place later this year: Support for the new GPT-4o that will help the company’s AI-powered content portal Box Hubs, support for additional file types including natural language queries on image file formats, and the ability for developers to extract “key information from documents at scale” via Box AI for Metadata API.

Company CTO Ben Kus said in a news release that the “combination of AI and unstructured data represents the biggest untapped opportunity in enterprise IT.”

Thomas Randall, director of AI market research at Info-Tech Research Group, said in an email that Kus is “right that the combination of AI and unstructured data is a large opportunity. What is missing from this statement is the solution’s ability to unify such data across different siloes. While organizations stand to benefit from optical character recognition for discovering and summarizing unstructured data, the real value is also ensuring the solution can discover and unify that data from across different systems.”

If an organization uses only Box for document storage, he said, “the problem is solved. If a company is using Box, SharePoint, Dropbox, or any number of other document management systems, organizations risk producing inconsistent business decisions based on disparate data.”

In terms of the overall launch, Randall said that “there is nothing important from a technology innovation perspective. Instead, the importance lies in Box needing to launch its own generative AI function to retain competitive market share. Already, organizations are starting to leverage smaller best-of-breed solutions that offer generative AI-driven knowledgebase and document management systems alongside document information extraction and unified business intelligence. In this context, Box will likely not experience a huge drive in new customer acquisition for their generative AI features alone.”

When it comes to generative AI and the organization, curation, and publishing of content, Amy Machado, senior research manager at IDC, said, “Everyone is still trying to figure it out. Not just the tech suppliers like Box, but also the end user organizations who are wondering, ‘How much value am I getting out of this, and how much am I willing to spend?’ It is not like we are in full production, and everybody knows how to price it, and everybody knows what is going to cost.”

The reality, she said, is that “we are in this kind of cool experimentation phase, and I think that having the query limit was putting pressure on end-users to really experiment if they had cost fears. It was a smart move by Box to eliminate the query cap for it allows enterprise customers to experiment and figure out how they can use this technology and get real value and benefit out of it.”

Box said that “access to GPT-4o for products such as Box Hubs, as well as well as support for new file types including images and spreadsheets in Box AI, is planned to be available later this year and will be included in Enterprise Plus plans.”

Meanwhile, Box AI for Metadata API is now in beta for customers on Enterprise Plus plans. The company said pricing will be announced closer to general availability, along with pricing for other Box AI platform API calls and end user metadata queries in the core Box application.

Randall said that the enhancements will be a welcome addition for “Box customers and prospects already considering Box. However, Info-Tech first recommends that organizations have an AI governance strategy in place to ensure that solutions like Box AI are utilized in appropriate contexts. Organizations need to review that the data Box AI is pulling from is not inconsistent, saving the solution from hallucinating or providing incorrect responses.”

An organization’s user base, he said, “should also be trained on prompt engineering and suitable use cases. The danger is that these tools are rolled out to an untrained workforce, who then become over-reliant on generative AI and lose sight of the proper use of the solution: as an assistant only.”