Keep pulling the thread on Arvind Jain.
Glean's platform bakes user permissions directly into its core indexing technology to ensure search results are always filtered by access rights.
To prevent data leakage in its custom models, Glean uses a curation process to train them only on a subset of enterprise data that is deemed safe.
Arvind Jain predicts that in the future, every knowledge worker will have a team of AI assistants, co-workers, and coaches that will perform 90% of their work.
AI agents can create "evergreen documentation" by synthesizing information from various sources like design docs, Jira tickets, and code commits to provide the most up-to-date view of a system.
A non-technical finance team member at Glean successfully used an AI agent to generate a customer churn risk report by analyzing data from Salesforce, Slack, and product dashboards.
To mitigate hallucinations, Glean's system verifies each line of an AI-generated answer against the source material provided in the prompt, suppressing or flagging information that cannot be traced back.
Arvind Jain believes the biggest challenge in enterprise AI is not model hallucination, but the retrieval of correct, up-to-date information from numerous disparate systems.
Arvind Jain is the CEO of Glean, which is described as one of the most successful enterprise companies foundationally using LLMs.
Before founding Glean, Arvind Jain founded Rubrik, which the speaker believes IPO'd at a valuation over $20 billion.
In early 2019, Google's search teams were already observing a significant positive impact from transformer technology on core search engine components.
Glean was likely the first company to apply transformer-based models to enterprise search in 2019.
Glean's technology strategy is to maximally leverage external innovations, such as large language models from Google, OpenAI, and Anthropic, rather than building its own foundational models.