Citi Gen AI Summit 2025 Takeaways: Fireside Chat with Arvind Jain, CEO of Glean

During the Citi Gen AI Summit, I had the pleasure of sitting down with Arvind Jain, the founder and CEO of Glean. Citi Ventures invested in Glean in April 2024 and we have enjoyed watching the company grow rapidly since then.

In our wide-ranging conversation, Arvind and I talked about his personal journey as an entrepreneur, how Glean has evolved its product offerings, and best practices for enterprise AI adoption and rollout.

Glean’s evolution and maturation

Glean has transformed from an enterprise search company into a versatile AI assistant and agent platform. This shift means Glean can now offer AI-driven question answering and conversational capabilities, similar to popular commercial models, but tailored for enterprise use across organizations globally.

This was not an accident. Arvind explained how he and his team were quick to respond to customer demand for custom applications, namely AI agents, powered by Glean’s underlying platform.

Glean has successfully become a central hub within enterprises for AI-driven productivity and knowledge management. Organizations are using its technology to create AI agents for all departments, such as HR, engineering, sales, and customer support — all within a single platform.

Build vs. Buy: advice for startups

Should startups and enterprises build their own AI models or leverage existing third-party ones? Arvind said it is too time-intensive and costly for startups to build their own models, and that foundation models are generally more capable than specialized ones, especially when it comes to quality and reasoning.

Glean prioritizes leveraging external models while focusing internal efforts on building semantic models trained on specific enterprise data. This allows Glean to incorporate the latest innovations from across the AI industry and securely deliver them to customers.

The primary goal for any AI application should be to ensure it works correctly, Arvind said, and that means using the best available models. In other words, startups should focus on their unique value-added features rather than trying to rebuild what already exists.

Successful enterprise adoption

Arvind cautioned against pursuing grand, overarching transformation ideas from the outset. Instead, he recommended focusing on small, manageable wins within individual departments. For example, in a legal team, Gen AI could be used to generate the first version of contract redlining, or in engineering, Gen AI could assist with initial code reviews.

Humans must be at the core of any enterprise Gen AI strategy, with the technology augmenting — not replacing — people. Arvind gave the following tips:

• To demonstrate ROI and build confidence in Gen AI, target specific tasks and workflows where the technology can provide immediate value
• Seek small wins, such as a 2× increase in efficiency with humans still fully in the loop
• Take a pragmatic approach and work within Gen AI's current limitations