Keep pulling the thread on Arvind Jain.
Arvind Jain predicts that in the future, every employee will have a personal team of AI assistants that proactively performs 90% of their work and acts as a coach for upskilling.
Glean's platform must enforce user-level permissions because approximately 90% of knowledge inside a company is private or has restricted access.
Some early potential customers were hesitant to adopt Glean because a powerful search tool exposed existing data governance gaps, such as unsecured documents containing employee salaries or M&A plans.
The idea for Glean originated from Arvind Jain's experience at Rubrik, where the company had 300 SaaS systems and he could not find a suitable commercial product to solve their internal search problem.
Glean was founded in early 2019 with the insight that transformer models could fundamentally change enterprise search.
Glean's product evolved from a Google-like search tool into Glean Assistant, a conversational AI that functions like ChatGPT but is grounded in a company's private data and knowledge.
The widespread adoption of SaaS applications with open APIs was the key technological shift that made a turnkey enterprise search product like Glean possible for the first time.
Most users find AI unintuitive because they have been trained by 20 years of using Google to type only one or two keywords into a search box.
One of Glean's largest customers has over 1 billion internal documents, a volume of content equivalent to the entire public internet in 2004.
In response to customer fears about data governance, Glean evolved to become a security product that helps companies fix governance issues and become "AI ready."
Large Language Models (LLMs) have fundamentally shifted the search paradigm from a static, keyword-based model to one that can deeply understand user questions and document content conceptually.
Arvind Jain believes that even with near-infinite context windows, LLMs will still require information to be pre-organized and structured to reason over it effectively.