The development of agent-to-agent protocols, such as A2A, could represent a fundamental shift in how businesses operate.

Box Inc.’s current artificial intelligence strategy is built on that premise of interoperability. The intelligent content management platform has partnered with Google Cloud to produce its own agents, which focus on data extraction and classification and workflow automation.

“Ninety percent of content that an organization has, whether it’s HR documents, whether it’s product documents, whether it’s financial documents, they’re all unstructured, which means that … the heart of a company that runs on content has been hard to access, hard to use” said Yashodha Bhavnani (pictured), head of AI at Box. “AI fundamentally just unlocks this.”

Bhavnani spoke with theCUBE’s John Furrier for the Google Cloud AI Agents in Action Series on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed A2A and Box’s collaboration with Google Cloud. (* Disclosure below.)

Transforming unstructured data into insights

Box has harnessed Google’s Gemini model and Vertex AI platform to develop its agents, including Box Extract, which performs detailed data extraction, and Box Automate, which manages business workflows. One of the company’s biggest goals is to make unstructured data feel structured, according to Bhavnani.

“If you had a 100-page lease agreement and you had multiple of those, and you wanted to drive in and say, ‘Well, when was this written? When does it end? What’s the value of the contract?’ instead of now parsing through 100 pages, the AI can extract that information for you,” she said.

Gemini’s multimodal capabilities help make this achievable. One of the use cases Bhavnani highlights is insurance claim files, which are proprietary and involve lots of unstructured data. Box Extract can analyze thousands of those claim documents and identify specific information, such as end dates and whether or not the claim was accepted

“They no longer have to sit there, look through 1,000 documents and make a spreadsheet,” Bhavnani said. “They can simply query because now these insights are attached to the documents securely in Box, and you can simply ask, ‘What were all the claims that were filed in the last month over the price range of X thousand?’ And that went from hours and hours of work to a few minutes, if that.”

The A2A future

Box sees itself as a “valuable link in the chain.” The future will not be single agents acting in silos, but agents working across the ecosystem. Having the A2A protocol with Gemini Enterprise is seen by Bhavnani as critical because it allows enterprises to use Box AI along with other agents in the cross-ecosystem.

“The best agents have the best context, and Box has that,” Bhavnani said. “Box has knowledge, all the knowledge of your entire enterprise. If you think about that, you can now use AI, build an agent, use agents with this context, and the richer that context is, the better the agents are.”

A2A is the next frontier in business workflows and is already being facilitated through the dedicated AI agent category on Google Cloud Marketplace, Bhavnani added. These platforms streamline reseller partnerships and the procurement of agentic AI products, enabling businesses to adapt agentic AI into their workflows with their unique architecture and costs in mind.

“When we offer Box AI agents for the Gemini Enterprise AI agent marketplace, what we’re really offering is we’re making it easier and a lot simpler for our customers to really capture the power of AI with their content while directly being in their own Google Cloud environments,” Bhavnani said.

The marketplace also enables enterprises to work with Box AI securely. A2A can only work if businesses trust the agents to interact with proprietary data in a safe manner. Even Bhavnani admits that working with AI isn’t all sunshine and roses, but she still believes the benefits greatly outweigh the costs.

“You have to have a sense of patience with AI,” she said. “I’ve seen magic where you build an agent — it’s awesome, it works the first time. But I’ve also seen that there is usually a sense of you have to just iterate through it, which is you’ll build the agent, ‘Oh, it didn’t give me that exact field that I wanted to iterate on with 90% confidence. How do I re-prompt it?’ So, there is a sense of patience, but it’s so worth it.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of the Google Cloud AI Agents in Action Series:

(* Disclosure: TheCUBE is a paid media partner for the Google Cloud AI Agents in Action Series. Neither Google Cloud, the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE