(US & Canada) | Building a Good Enterprise Search Product Requires Understanding People — Glean Founder and CEO

Arvind Jain, Founder and CEO of Glean, speaks with Kirk Ball, EVP and Chief Information Officer at Giant Eagle, in a video interview about Glean as a product, the architectural and technical layers in it, the role of LLMs in Glean, successful use cases, the future of work with generative AI, and the company’s prospects.

Glean is an AI platform for enterprise search and knowledge discovery.

Jain introduces Glean as Google and ChatGPT for the enterprise. It connects with all the enterprise data and answers employee questions, leveraging internal company knowledge. 

When asked what fueled the idea, Jain stresses the challenge of finding scattered information. Recalling his previous job at Rubrik, he shares how a survey revealed that the work was not progressing because employees did not know where to look to find information.

With businesses having massive information spread across systems that grow exponentially every year, people should be able to find the required information easily, says Jain. That is when he realized the need to create something that would help people find things at work, and Glean was created in 2019.

Delving further, Jain says that the idea of solving the problem came from his experience working on Google Search. He adds that the founding team included former Google Search engineers who worked over a decade to build Search.

The first part of building the team was finding people with the right experience and utilizing those while building Glean, says Jain. He affirms that solving the enterprise search problem was different from solving the web search problem, but there were relevant experiences that could be used.

Sharing one such experience, Jain says building a good search product requires understanding people, their information needs, and how they engage with web or enterprise knowledge.

Speaking of the architecture and technology approach of Glean, Jain states that it connects with all the different enterprise needs, and there is a rich library of deep integrations.

Glean first connects and brings data from all systems into its unified search index. Then comes the governance aspect, because enterprise information needs to be governed. This forms the second part of the core technology, which revolves around understanding permissions, governance, and access controls.

Adding on, Jain affirms that the company keeps track of the access part while indexing information. The third part of the architecture is understanding the enterprise and building the knowledge graph, which includes people, the actual data, and the relationships between them.

Highlighting knowledge, Jain says that it is also about taking a document and building an understanding of what it is about, who uses it, who created it, and what purpose it serves. This knowledge graph is actually essential to rank and answer questions correctly for people.

In continuation, Jain states that when asked a question, Glean will know who the user is, use the knowledge that one is authorized to know, and display results accordingly. Also, it can personalize the deliverable information based on the employee’s role and their knowledge of the enterprise.

However, this is where AI models come into play. Jain confirms using LLMs before they became popular. He notes that an LLM understands the conceptuality of the question and the semantics of the document and matches those.

According to Jain, semantic search or vector search is a big part of Glean, and it is done by training the models on the enterprise corpus. The company is leveraging the power of language models to rank well and surface the best information on top.

In addition to that, LLMs are utilized to understand people’s intent and questions, and apart from furnishing just the information, with AI, one can synthesize the exact information that is needed.

Sharing some use cases, Jain first mentions that one of the largest global telecom companies is using Glean as the core tool available for over 100,000 customer care agents. It helps the customer care agents by putting the right information in front of them and using the knowledge base. Consequently, there has been a 25% reduction in case resolution time with that.

The second use case involves one of the largest engineering companies which deployed Glean across the company as an open-ended general assistant to everybody. For instance, it helps engineers with troubleshooting issues. As a result, engineering teams have reported 145 engineering years’ worth of time savings.

Moving forward, Jain states that many companies, such as AWS, have well-documented articles on how to use their services, but they must be searchable. Therefore, companies are putting customer-relevant information on their platforms and are working to build a great search product where customers can talk to the knowledge base to solve problems.

When asked how generative AI will change the future of work, Jain states that in 3–5 years, most of the manual assisting work will be done by AI. With that in mind, Glean plans to launch an AI-based assistant for everybody working in the enterprise.

Having an assistant that schedules meetings, manages the calendar, replies to emails, and understands needs is a luxury only available to executives. But in the future, all employees should have a smart AI assistant taking mundane work off their plates.

Furthermore, Jain maintains that although AI is powerful, it is challenging to tap into its potential. Emphasizing hallucinations, he says that the technology is not predictable, and organizations have to work hard to make it reliable. Moreover, it is potentially dangerous, and companies must ensure delivering a safe, secure, and responsible AI experience.

Additionally, organizations must address the gaps in governance because sensitive information is not protected adequately. He urges enterprises to utilize technology and find the right products that help deliver safe AI across enterprises.

Commenting on Glean’s future, Jain states that the company is still in its early stages and aspires to deliver “magic” by having all the right answers. He says that getting the right answers takes time, and the company will keep working on it to make it better, more accurate, and more reliable.

The company will not be building more products because it is focused on making Glean more capable. However, one interesting addition will be bringing in AI assistance for all.

Concluding, Jain asserts that in terms of assistance, the system has to be proactive, and the company is directed towards building that. The company envisions having Glean as the de-facto tool used by every single worker in every single company to enhance productivity.

CDO Magazine appreciates Arvind Jain for sharing his insights and success story with our global community.