Google aims to boost productivity with AI Agent tool

Employee productivity

Google Agentspace seeks to give employees a single place to search for information across the enterprise, instead of searching across multiple tools, Google said.

Employees may spend hours searching for internal company information because the information is in the form of unstructured data.

"Agentspace combines the best of search and AI agents," said Raj Pai, vice president for product management at Google, during a press briefing about the new product.

The strength of Agentspace is helping employees make sense of the digital pile of information within their organization, said Paul Baier, CEO and co-founder of GAI Insights, a GenAI analyst firm.

Reasoning and multimodality

Google seeks to accomplish two goals with Agentspace: bring better reasoning capabilities and include multimodal capabilities, said Gartner analyst Arun Chandrasekaran.

Agents need better memory, which is the ability to remember things and contextualize the responses.

"[AI] agents need better reasoning," Chandrasekaran said. "Agents need better memory, which is the ability to remember things and contextualize the responses."

Google accomplishes the goal of better reasoning through Gemini and its history with search.

"Search can take advantage of the planning, the reasoning and the memory that agents provide to answer complex questions," Pai said. "Agents are able to ensure that all generated content is grounded in complex enterprise data."

Google establishes multimodality by including built-in multimodal search capabilities across text, video, images and audio in Agentspace. It also allows users to generate audio summaries through the NotebookLM audio feature.

Google's inclusion of a built-in translation feature is also beneficial for users, Chandrasekaran said.

"Google being Google, they're able to support so many more languages than others can," he said. "Thanks to the Android ecosystem, they're also able to train these models well using a lot of that data."

The use cases Google mentioned for Agentspace include a marketer generating content for a new product release or an HR department helping employees onboard faster.

"It's always good to start with kind of more inward, focused use cases," Chandrasekaran said. "Given that agents are super new, and people are concerned about things like 'Oh, I don't want the agents to make a decision without consulting me.'"

Compared with Salesforce Agentforce and Microsoft Agents, Google's Agentspace seeks to help developers create their agents while integrating agents into Google Cloud services, Chandrasekaran continued.

AI agent challenges

A key challenge for Agentspace is having employees learn to use the tool, Baier said.

"The biggest initial weakness is not in the tool but in training employees on how to best leverage and make Agentspace a part of their daily, personal workflow," he said.

Another challenge will be how Google handles responsible AI questions regarding agents, Chandrasekaran said.

"What are they doing to demystify, like making these agents more explainable, more interpretable, or preventing egregious biases or toxic content coming out of these agents," he said.

He added Google also needs to have the right underlying data infrastructure for agents to flourish.

"There's a lot of work that needs to be done there, and it'll be very interesting to see how Google does that more holistically moving forward," he continued.

Google said it's putting effort into partnering with customers to understand the various data.

Esther Shittu is a TechTarget Editorial news writer and podcast host covering artificial intelligence software and systems.