As search technology has evolved, generating a list of local top-rated bakeries, to name a common use case, has become a normal part of life. Yet the same capabilities for finding obscure pieces of information or insight from a company’s vast database are not so easily delivered.

That’s the challenge involved in enterprise search, accessing critical information across billions of sensitive, proprietary files within an organization with the same speed and accuracy that people are accustomed to for searches on the web. It’s also the business model for Glean Inc., founded in 2019 by a team of former Google engineers and seasoned technologists with a mission to transform AI-powered enterprise search.

On Tuesday, Glean announced general availability and expansion of its agentic AI platform. The latest releases were designed to empower employees to build, scale and deploy AI agents.

“The enterprise brings a lot of messiness, a lot of complex permissions,” said T.R. “Vish” Vishwanath, co-founder and chief technology officer at Glean. “There’s a lot of sensitive content. The AI transformation in the enterprise will take time. The first step is to make everyone powerful.”

Teams of agents for personalized support

Vish spoke with SiliconANGLE in an exclusive interview as part of a press briefing this week in San Francisco. At the GleanGo conference on Tuesday, the company announced that it would provide customers with an open, horizontal agent environment powered by the firm’s search and reasoning engine.

“Our vision is that you will have a team of agents,” Vish said. “They can uplevel your work.”

This team concept was underscored with the launch of a library of more than 30 prebuilt quickstart agents with capabilities across engineering, sales, human resources and information technology that can be customized and deployed for use cases such as ticket resolution, prospecting, code review and help desk automation.

The intent is to create a work environment where each employee can have personalized support, much like what interns have historically provided in organizations, Emrecan Dogan, vice president of products at Glean, said in a conversation with SiliconANGLE.

“All of a sudden instead of an employee waiting for years to get their first intern; now every employee can overnight get their team of interns,” Dogan said. “We believe there’s an opportunity to craft AI so that employees do their best work.”

Bridging the knowledge gap

For employees to do their best work, they must first understand how to leverage the technology to perform key tasks. This has been a clear imperative for Glean. As part of its release this week, Glean has included new integrations with Snowflake Inc. and Workday Inc. to facilitate the use of agents.

Snowflake Cortex Analyst enables users and Glean’s own agents to query Snowflake-stored data using natural language. The Workday interface allows secure, bidirectional interoperability between Glean’s agents and Workday’s to act on information across both platforms.

“The number one challenge that businesses are facing today is how to make sure the workforce is learning this new technology,” founder and Chief Executive Arvind Jain (pictured) said during a question/answer session with the media this week. “That’s the first challenge that has to be solved. It’s very hard because people are busy.”

Glean has also built its reputation around a robust reasoning engine. The company’s agents are built and trained to understand important enterprise data points. This can be a complicated task in an organization with numerous product lines run by teams of people with job responsibilities that are not often clear.

“The best way to describe what Glean does is perform as a system of context for the enterprise,” said Dogan. “It understands your data. AI experiences win, they have amazing context. Connect that meeting to that project with the right team member.”

Building protections and guardrails

What if the wrong team member is selected and provided with sensitive company information they didn’t have permission to access? To guard against this possibility, the company has introduced Glean Protect, permissions infrastructure that provides granular access controls that manage who can run, view or edit agents in the enterprise.

“There’s no possibility that an agent will give you access to information you don’t have access to,” said Thai Tran, product lead for Glean’s AI assistant in an interview with SiliconANGLE.

Glean’s governance protocols extend to protection against AI-specific threats. The Glean platform includes built-in safeguards against prompt injection and jailbreak attempts, where hackers will seek to bypass security measures around a large language model. Security is grounded in Glean’s ability to have a complete picture of an enterprise customer’s data stores, according to Sunil Agrawal, chief information security officer at Glean.

“We will shine the light on where all sensitive information is placed,” Agrawal said in an interview with SiliconANGLE. “It’s not just broad, we go deep as well. If you have a billion documents in Google Drive, we will tap into all of those. We don’t do sampling.”

Over the course of its six-year history, Glean has managed to build a roster of major enterprise customers. Clients include Databricks Inc., Deutsche Telekom, Pure Storage Inc. and the data streaming platform Confluent Inc.

Along the way, Glean transitioned its business model from enterprise search into applications that leverage generative AI and autonomous agents. That has allowed the company to reach more than $100 million in annual recurring revenue and a valuation of more than $4 billion.

The company’s growth has been based on an ability to navigate the complexities that surround information stored in enterprise files, and this is powering the expansion of its agentic AI platform.

“To be an assistant you need to be able to understand people’s data,” said Tamar Yehoshua, president of product and technology at Glean. “This is the core of what Glean does. We want to be the best at finding answers from any question, any information.”

Photo: Mark Albertson/SiliconANGLE