Box today previewed four new agentic AI tools for search, research, metadata extraction and connectivity to Microsoft apps.
The Search agent uses generative AI to return results across large swaths of documents. The Deep Research agent can synthesize complex findings from files into content such as research reports or trend analyses. For example, a legal team could examine indemnification clauses across different clients' contracts, or insurance auditors could analyze terms and payout limits for thousands of policies at once.
The third agent, Enhanced Data Extraction, adds structured metadata to many different files, including PDFs and even handwritten notes. A fourth, AI Agent for Microsoft 365 Copilot and Chat, connects a user's Microsoft app content in Box with Copilot AI.
The agents will be available in the coming months, with pricing to be determined closer to the general availability dates, according to Box.
Like many companies, including Microsoft, Oracle, Salesforce and ServiceNow, Box is getting into the agentic AI game -- but it's still a technology in its infancy, said Alan Pelz-Sharpe, founder of Deep Analysis, an independent research firm.
All those companies are now releasing agents with basic functionality, in preparation for future years. Then, these agents can be combined to automate the more complex business processes that humans are doing now, he predicted.
The vast majority of agents today are just relabeled bots -- there's no autonomy there, they're not real agents.
Alan Pelz-SharpeFounder, Deep Analysis
"The vast majority of agents today are just relabeled bots -- there's no autonomy there, they're not real agents," Pelz-Sharpe said. "That doesn't mean it's not real. It's just that it's very early days, [when] you pick the low-hanging fruit."
Pelz-Sharpe said the significant upgrade to Box's previous AI search and metadata tools will be the addition of Box's reasoning engine.
Box co-founder Aaron Levie said the reasoning engine can understand more detailed queries and go deeper into enterprise documents to draft contracts or marketing materials.
Not only are these agents available within Box, but the company has also made them available for joint customers of numerous platforms, including Salesforce, ServiceNow, AWS, Adobe and Google.
"We want to make the entire platform intelligent and ensure that you can effectively interact with your data from any other system in an AI-first way," Levie said.
"So if the user is working from ChatGPT, we want to show up in ChatGPT. If the user is working in Claude from Anthropic, we want to show up in Claude," he said. "A user can say, 'I want you to go run analysis on this set of content, the content's in Box, but I want the result to show up in this interface that I'm working in right now.' That's the future end state of a lot of these capabilities."
Box AI Agent for Microsoft 365 Copilot can scan Microsoft app documents for snippets of data, such as contract expiration dates, and return results within Copilot.
Some of the more ambitious use cases for enterprise content agents, such as those that can author reports by themselves, will still likely need a human in the loop for review and verification, Pelz-Sharpe said. Some of the agents coming out now -- including Adobe's -- are better than others. And since Box's Deep Research agent isn't out yet, it remains to be seen how much of a head start it will give users on big content projects.
But even the best content agents will still require humans to check AI drafts closely, he said.
Box previewed its agentic AI in conjunction with its May 15 Content + AI virtual summit.
Don Fluckinger is a senior news writer for Informa TechTarget. He covers customer experience, digital experience management and end-user computing. Got a tip? Email him.
