Arvind Jain: Why Now Is the Time to Solve Enterprise Search (Encore)

Michael Mignano welcomes listeners back to Generative Now for an encore presentation featuring Arvind Jain, founder and CEO of Glean. The conversation explores one of the most sophisticated AI workplace assistants that has become a massive player in the enterprise search space. Arvind brings impressive credentials as an alum of Microsoft and Google, and previously co-founded Rubric, which he successfully took public in the high-stakes world of data security.

The discussion promises to cover the early days of Glean, how AI is shifting the future of work, and what's next for the company. What makes this story particularly fascinating is that Glean began well before the current generative AI explosion that started with ChatGPT's launch, suggesting Arvind saw opportunities that others missed.

Timestamp: [0:00-1:18]

Arvind reveals that Glean originated from a painful problem he experienced firsthand at Rubric. While the company was growing rapidly and achieving success, they encountered a troubling pattern as they crossed the thousand-employee threshold: dramatic productivity drops across all teams. This wasn't the typical organizational complexity that comes with growth—it was something much more severe.

The root cause became clear through employee pulse surveys: people couldn't find the information they needed to do their jobs effectively. Company knowledge was fragmented across multiple systems, and as the organization grew quickly, it became unclear who worked on what and who could provide expertise on specific topics.

"People were saying that you know I'm not enabled I can't actually get my work done because of you know these these issues" - Arvind Jain"

This was the moment Arvind realized that despite his decade of experience building Google search, there was no good enterprise search product available to solve this fundamental workplace problem.

Timestamp: [1:18-3:37]

What set Glean apart was Arvind's early recognition of a technological inflection point. In 2018, before the current AI wave, search engineers were already witnessing the power of transformer-based neural networks. Models like BERT were demonstrating capabilities that would revolutionize search technology.

From his experience at Google search, Arvind saw how transformer-based models could rebuild every component of search technology and outperform systems that had been developed over the previous 15 years. The key insight was that transformers enabled search to operate at a much more conceptual and semantic level, rather than just keyword matching.

"Glean became the the first company I believe you know that has brought the transformer technology to the enterprise" - Arvind Jain"

This technical foresight, combined with the urgent business problem he had witnessed at Rubric, created the perfect opportunity to build something fundamentally new in enterprise search.

Timestamp: [3:37-4:56]

Michael shares his own experience of joining a large company and facing the challenge of disconnected information across Slack, email, Google Drive, Confluence, and other systems. This resonates with the core problem Glean set out to solve—bringing together fragmented enterprise knowledge.

Arvind explains the foundational technical approach: since most enterprise knowledge now lives in SaaS cloud-hosted systems that interoperate well through APIs, Glean began building integrations to access this distributed information. The key insight was that modern enterprises operate in a SaaS-first world where information is naturally distributed, but there was no unified way to search across these systems.

The technical stack required not just connecting to these systems, but understanding how to search across them effectively. This is where the transformer technology became crucial—enabling semantic understanding rather than simple keyword matching across diverse data sources.

Timestamp: [4:56-6:04]

• Glean started in 2019, well before the ChatGPT-driven AI boom, showing prescient timing on enterprise AI needs
• The founding insight came from experiencing severe productivity drops at Rubric when the company hit 1,000+ employees
• Employee surveys revealed the core problem: inability to find information and expertise within the organization
• Arvind leveraged his decade of Google search experience to recognize that no good enterprise search solution existed
• Transformers were already showing promise in search applications by 2018, before they became mainstream
• Glean was among the first to bring transformer technology to enterprise applications
• Modern enterprises operate in a SaaS-distributed world that creates natural fragmentation of knowledge
• The technical approach focused on API integrations to connect cloud-hosted enterprise systems

Timestamp: [0:00-6:04]

Companies:

• Rubric - Arvind's previous company, an enterprise data security company that he co-founded and took public
• Google - Where Arvind spent over a decade building Google search technology
• Microsoft - Arvind's background includes experience at this tech giant
• Lightspeed - Michael Mignano's venture capital firm

Technologies:

• ChatGPT - Referenced as the catalyst for the current AI explosion
• BERT - Family of transformer-based language models that were early examples of the technology
• Transformers - Neural network architecture that Glean recognized early as revolutionary for search

Platforms/Systems:

• Slack - Enterprise communication platform mentioned as one of disconnected information sources
• Google Drive - Cloud storage platform mentioned as containing fragmented company knowledge
• Confluence - Enterprise wiki platform mentioned as another disconnected information source
• Octa - Identity management platform mentioned in the enterprise tool ecosystem

Timestamp: [0:00-6:04]

Arvind explains the fundamental technical architecture behind Glean's approach to enterprise search. The first critical step was gaining access to information across various enterprise systems. Since modern enterprises operate primarily on SaaS cloud-hosted systems that interoperate well through APIs, Glean built integrations with products like Confluence, Jira, Google Drive, and SharePoint using their published APIs.

However, enterprise search presents unique challenges that web search doesn't face. Unlike the open internet where all information is accessible, enterprise information is privileged and secure. Certain documents are accessible to some employees but not others, creating a complex permission landscape.

"You have to actually build a safe version of search you can't actually start to leak information to people uh from this search experience" - Arvind Jain"

This required building extensive security infrastructure to handle permission-aware access within search results, ensuring that the search experience respects existing access controls across all integrated systems.

Timestamp: [5:46-7:17]

Contrary to what might be expected, Glean incorporated transformer models from the very first version of their product. The technology was powerful enough and available enough that they could build much smarter search capabilities from the beginning. Users could ask questions in natural language and receive conceptually and semantically matched results, rather than relying on brittle keyword-based search.

Glean's approach was sophisticated: they took BERT, an open-domain model that Google had made available, and customized it for each enterprise customer. For every client, they would retrain these models on the company's specific data corpus in a secure way, enabling the model to understand company-specific terminology, lingo, code names, and acronyms.

"We would actually train we would actually train you know retrain these models on the enterprise corpus on that data in a safe and secure way again um so that you know the the model starts to understand your company your lingo your code names you know acronyms you know all of that stuff" - Arvind Jain"

The core search stack used a hybrid approach, combining traditional search techniques with transformer-based semantic matching technology, providing the best of both worlds from day one.

Timestamp: [7:17-8:56]

Michael draws a comparison to the magical first-time experience that many AI startups create today, wondering if Glean had that same impact when introducing transformers to enterprises in 2019-2020. Arvind reveals an interesting perspective: they didn't actually market transformers as the cool technology.

Instead, the appeal was practical and immediate. Users could ask questions in natural language and get conceptually matched results without needing to remember exact keywords. This solved a major pain point with traditional search systems where people knew the information existed (like an email in Outlook) but couldn't retrieve it because they couldn't recall the precise words.

"I think what they cared about was what you could do with it which was that well like you know in glean you can ask questions in natural language and it will conceptually match the information with your question" - Arvind Jain"

The transformer technology enabled this natural interaction, but the focus was on the user experience improvement rather than the underlying AI sophistication.

Timestamp: [8:56-10:14]

Despite the obvious utility of their solution, Arvind candidly admits they didn't achieve product-market fit for a long time. The fundamental challenge was that the industry viewed enterprise search as a "vitamin" rather than a "painkiller." While everyone acknowledged the problem of fragmented information across systems and the struggle to find things, companies felt they were managing adequately.

There was no existing category of search products that enterprises were actively purchasing, which meant Glean had to create an entirely new market. This required extensive evangelism and education about the importance of solving search problems.

"We had to create a market and it took us time we had to do a lot of evangelism we had to talk about why it is important like you know it reduces frustration it makes your employees happier you know they spend a third of their time just looking for information so you'll save a lot of time" - Arvind Jain"

The value proposition centered on reducing employee frustration, improving happiness, and saving significant time—since employees spend roughly one-third of their time searching for information.

Timestamp: [10:14-11:22]

When asked about the moment everything changed, Arvind identifies the first major inflection point: reaching approximately 30 companies, with a strategic focus on the technology sector. Getting the most iconic tech companies to use Glean at scale became crucial for generating authentic word-of-mouth marketing.

This tech-first approach proved strategic because technology companies would naturally be more receptive to innovative AI-powered solutions and could serve as influential reference customers for other enterprises. Once these prominent tech companies validated the product and began experiencing its benefits, organic advocacy began to emerge.

The focus on quality over quantity in these early customer relationships created a foundation for sustainable growth through genuine customer satisfaction rather than just marketing efforts.

Timestamp: [11:22-11:54]

• Glean built integrations using published APIs from SaaS platforms, taking advantage of the cloud-first enterprise environment
• Enterprise search requires sophisticated security infrastructure to respect permission-based access controls
• Transformer technology was implemented from day one, not added later during the AI boom
• Each enterprise customer received a custom AI model trained on their specific data and terminology
• The hybrid search approach combined traditional techniques with semantic matching for optimal results
• Marketing focused on practical benefits (natural language queries) rather than underlying AI technology
• Enterprise search was initially viewed as a "vitamin" rather than a "painkiller," requiring market creation
• Companies spend approximately one-third of employee time searching for information
• Product-market fit required extensive evangelism and education about the value of enterprise search
• Strategic focus on tech sector customers created influential reference cases and organic word-of-mouth

Timestamp: [5:46-11:54]

Technologies:

• BERT - Google's open-domain transformer model that Glean used as the foundation for their custom enterprise models
• Transformers - Neural network architecture that enabled natural language search capabilities

Enterprise Platforms:

• Confluence - Atlassian's enterprise wiki platform integrated by Glean
• Jira - Atlassian's project management tool integrated by Glean
• Google Drive - Google's cloud storage platform integrated by Glean
• SharePoint - Microsoft's collaboration platform integrated by Glean
• Outlook - Microsoft's email platform referenced as an example of search challenges

AI Models:

• ChatGPT - Referenced as an example of magical first-time AI experiences
• Image generators - Referenced as examples of powerful AI tools with immediate impact

Timestamp: [5:46-11:54]

Arvind identifies two distinct moments that transformed Glean's trajectory. The first came when they reached approximately 30 companies, strategically focusing on the technology sector. Getting the most iconic tech companies to adopt Glean at scale created powerful word-of-mouth momentum and generated organic inbound interest from other enterprises.

The second transformative moment was the launch of ChatGPT, which fundamentally changed how enterprise leaders viewed AI capabilities. This created an unexpected but powerful marketing effect for Glean.

"In some ways you can think of you know glean as you know the enterprise version of chat GBT like you know it does everything that chat GBT does but it does it like you know with that knowledge and context of your company" - Arvind Jain"

When enterprise leaders witnessed ChatGPT's capabilities, they immediately began imagining what it would be like to have similar AI power but trained on their internal company data, employees, and processes. This realization created organic demand and represented a massive second wave of momentum for Glean.

Timestamp: [11:35-12:52]

Michael astutely observes that ChatGPT essentially performed external marketing for Glean, validating the concept of conversational AI while highlighting Glean's unique enterprise advantages. All the foundational work Glean had done on security, safety, and permission management became crucial differentiators when bringing transformer technology into enterprise environments.

This represents a perfect example of timing and preparation meeting opportunity. While ChatGPT demonstrated the power of conversational AI to the world, Glean had already solved the complex enterprise-specific challenges that would prevent companies from simply adopting consumer AI tools.

The years of building secure, permission-aware systems suddenly became Glean's competitive moat in a world where everyone wanted enterprise AI but few could deliver it safely.

Timestamp: [12:52-13:10]

When asked about model management and whether Glean trains its own models or swaps between providers like OpenAI and Anthropic, Arvind clarifies their architectural philosophy. Glean is explicitly not a foundation model company and doesn't train the massive models that power general AI capabilities.

Instead, Glean's architecture resembles ChatGPT from a user experience perspective—users ask questions and receive intelligent answers. However, the underlying system can draw from both global knowledge and internal company data to provide responses. The challenge is that none of the major foundation models (GPT, Claude, Gemini, DeepSeek, etc.) are trained on any specific company's internal information.

"None of those models are trained on your enterprise information like so they don't know anything about like you know how work happens inside your company" - Arvind Jain"

This fundamental limitation of foundation models creates the opportunity for Glean's specialized approach to enterprise AI.

Timestamp: [13:10-14:14]

Arvind explains Glean's core technical architecture, which follows the Retrieval-Augmented Generation (RAG) pattern that has become standard for enterprise AI applications. This is a two-step process that solves the fundamental problem of making foundation models work with company-specific information.

First, when a user asks a question, Glean uses its retrieval and search technology to assemble relevant pieces of information from within the company's data ecosystem. Second, all of this contextual information is provided to the foundation model, enabling it to reason over the company-specific data and generate appropriate answers or responses.

"It's a two-step process like you know given any task or any question that a user comes and asks you know first we use our code retrieval and search technology to assemble like some relevant pieces of information from inside our company and then we give all of that to the model so that the model can actually reason over it" - Arvind Jain"

This RAG-style architecture has become the typical approach for most enterprise AI agents and applications, with search/retrieval as the first stage and model reasoning as the second stage.

Timestamp: [14:14-15:04]

Glean's architecture employs a sophisticated hybrid approach to model usage. They continue to build and train smaller models based on BERT or modern BERT-like architectures—what the industry now calls Small Language Models (SLMs). These custom models are trained on each enterprise's specific corpus and are primarily used for semantic matching of user queries with enterprise knowledge during the retrieval phase.

For the reasoning and generation phase, Glean leverages foundation models from various providers. Importantly, their architecture isn't tied to any single model provider, allowing them to optimize for different use cases.

"Our architecture is like we're not tied to one particular model like what we're seeing in the industry is that um there are many many model providers they're all in different models are getting better at different things" - Arvind Jain"

This flexibility is crucial because different models excel at different tasks. For example, Claude is currently considered superior for code generation compared to other models. By maintaining model-agnostic architecture, Glean can leverage the best capabilities from different providers for different use cases.

Timestamp: [15:04-15:58]

• Two key inflection points: reaching 30 tech companies for word-of-mouth, and ChatGPT's launch creating enterprise AI demand
• ChatGPT served as external marketing for Glean by demonstrating conversational AI capabilities to enterprise leaders
• Glean's early investment in security and permissioning became a crucial competitive advantage
• Foundation models aren't trained on enterprise data, creating the fundamental need for specialized enterprise AI
• RAG architecture has become the standard for enterprise AI: retrieval first, then model reasoning
• Glean uses hybrid approach: custom SLMs for semantic matching, foundation models for reasoning
• Model-agnostic architecture allows leveraging different providers' strengths for different use cases
• Different foundation models excel at different tasks (e.g., Claude for code generation)
• Enterprise AI requires solving problems that consumer AI tools cannot address safely

Timestamp: [11:35-15:58]

AI Models:

• ChatGPT - OpenAI's conversational AI that created market awareness and demand for enterprise AI
• Claude - Anthropic's AI model, noted as superior for code generation
• GPT - OpenAI's foundation model family
• Gemini - Google's foundation model
• DeepSeek - AI model provider mentioned in the foundation model landscape
• BERT - Google's transformer model that Glean continues to use as basis for custom enterprise models

Technical Concepts:

• RAG (Retrieval-Augmented Generation) - The two-step architecture combining search/retrieval with model reasoning
• SLMs (Small Language Models) - Industry term for smaller, specialized models like those Glean trains for semantic matching
• Foundation Models - Large general-purpose AI models that power reasoning and generation

Companies:

• OpenAI - AI company behind ChatGPT and GPT models
• Anthropic - AI company behind Claude

Timestamp: [11:35-15:58]

Arvind elaborates on Glean's model-agnostic approach, explaining how different foundation models excel at different capabilities. GPT performs better for reasoning tasks, while Claude leads in code generation. Rather than forcing customers to choose specific models, Glean's architecture automatically leverages the best model for each particular task.

"Our goal you know at glean is to make sure that for our enterprise customers like we can bring all the amazing innovation that's happening in the industry um you know to our customers they don't they should not have to choose you know what model to use" - Arvind Jain"

This approach means that as the AI industry continues to rapidly improve—with models getting better seemingly every week—Glean's product automatically benefits from these advances. The underlying architecture provides access to all different models and intelligently selects the optimal one based on the specific enterprise task at hand.

Timestamp: [15:58-16:41]

Michael highlights an exciting aspect of Glean's model-agnostic architecture: as foundation models continue to improve rapidly, Glean's product automatically gets better without requiring internal development effort. Any new capabilities that emerge from model provider innovations immediately become available to Glean's enterprise customers.

This creates a particularly compelling dynamic where model improvements unlock entirely new feature sets for Glean. As models become truly agentic and evolve from individual information retrieval to sequential, task-oriented workflows, this opens up massive new service areas for both Glean and their customers.

"As these models become truly agentic and they move from you know they move from individual sort of retrieval of information to more sequential task oriented things like that like that's got to up that's got to open up a huge service area for for Glean" - Michael Mignano"

The compound effect means Glean benefits from the entire AI industry's R&D investments while focusing their own efforts on enterprise-specific challenges.

Timestamp: [16:41-17:28]

Arvind confirms that new agentic features are arriving constantly, citing the recent introduction of "operators" as a game-changing development that dramatically expands Glean's ability to work with enterprise applications. These capabilities represent a fundamental shift in how AI can interact with business systems.

The vision is seamless data access for business users. If someone has a question or task requiring enterprise data, AI should make the process effortless. Previously, users might know data existed but found accessing it difficult. Now, through integrations and new LLM capabilities like computer use and operators, AI can interact with systems even without APIs by directly controlling browsers and interfaces.

"AI is going to make it super easy like you know a lot of it you know we do through our integrations uh but a lot of it we can also do through these you know like new LLM capabilities like computer use you know operators where you don't even need those systems to have APIs" - Arvind Jain"

This eliminates the traditional barrier where system integration required extensive API work, opening up possibilities for working with any enterprise system.

Timestamp: [17:28-18:40]

Arvind articulates Glean's core strategic philosophy regarding the broader AI ecosystem. Rather than competing with model providers and cloud hyperscalers, Glean deliberately leverages all the technology being built by these major players. Their focus remains on solving the enterprise-specific challenge: making advanced AI technology work seamlessly with company data and processes.

"Everything that the model providers that the cloud hypers scale retailers all the technology that they're building you know we we basically don't compete with them you know we leverage that technology and and uh and we do the remaining part which is like you know how do you make it easy to access that like you know all that technology and make it work on your data in your enterprise" - Arvind Jain"

This strategy allows Glean to benefit from the massive R&D investments of Big Tech while focusing their engineering efforts on the unique challenges of enterprise AI implementation, particularly around security, permissions, and data integration.

Timestamp: [18:40-19:02]

Arvind reveals that Glean now offers three distinct but integrated capabilities. First, Google-like search functionality for enterprise information. Second, a ChatGPT-like AI assistant that can answer questions and provide guidance. Third, and perhaps most transformative, an agent building platform that enables users to create custom agents for transforming business processes.

This agent building capability represents a democratization of automation and AI application development. Business users no longer need to be developers to build complex, interesting solutions. Many agents can be created simply by expressing business processes in natural language to the AI system.

"The best thing with AI is that it's actually bringing that power to you know to every business user you don't have to be a developer anymore to be able to build something you know complex or like interesting" - Arvind Jain"

The AI can take these natural language descriptions, apply reasoning, and create multi-step workflows that users can review and execute, fundamentally changing how work gets accomplished in enterprises.

Timestamp: [19:02-20:16]

When Michael asks about the technical requirements for agent building—whether it needs models specifically capable of reasoning, web browsing, or computer use—Arvind clarifies that it's a combination of factors. Working with just an LLM alone is insufficient for building agents that can automate business processes effectively.

The complexity of real enterprise automation requires more than language model capabilities. It demands integration with multiple systems, understanding of business logic, workflow orchestration, error handling, and the ability to interact with various interfaces and data sources.

"Today like you know if you just work with an LLM uh that's not enough for you to actually build an agent and automate a business you know business process with it" - Arvind Jain"

This insight reinforces Glean's value proposition: while foundation models provide the reasoning capabilities, enterprise AI requires sophisticated orchestration, integration, and workflow management that extends far beyond what any single model can provide.

Timestamp: [20:16-20:58]

• Different foundation models excel at different tasks (GPT for reasoning, Claude for code generation)
• Model-agnostic architecture allows automatic optimization for specific enterprise tasks
• Glean's product automatically improves as foundation models advance, without internal development effort
• New agentic capabilities like operators and computer use dramatically expand enterprise integration possibilities
• AI can now interact with systems without APIs by directly controlling browsers and interfaces
• Strategic focus on leveraging rather than competing with Big Tech AI investments
• Three core products: enterprise search, AI assistant, and agent building platform
• Agent building democratizes automation, allowing business users to create complex workflows without coding
• Natural language can be used to express business processes that AI converts into executable workflows
• Simple LLMs alone are insufficient for enterprise automation - requires orchestration and integration capabilities

Timestamp: [15:58-20:58]

AI Capabilities:

• Operators - New LLM capability that significantly expands enterprise application integration
• Computer Use - AI capability allowing direct browser and interface control without requiring APIs
• GPT - Referenced for superior reasoning capabilities
• Claude - Referenced for superior code generation capabilities

Technology Providers:

• Cloud Hyperscale Providers - Major cloud computing platforms that Glean leverages rather than competes with
• Model Providers - Companies developing foundation models that Glean integrates

Product Comparisons:

• Google - Referenced for search functionality comparison
• ChatGPT - Referenced for AI assistant functionality comparison

Timestamp: [15:58-20:58]

Arvind explains that building effective enterprise agents requires more than just working with LLMs. While LLMs provide foundational capabilities, successful automation of business processes demands additional infrastructure and orchestration. The key challenge is bringing the right enterprise data to these models in a secure and contextual manner.

Computer use represents an evolution in how LLM providers are expanding capabilities—it's not just a model improvement but a feature that enables models to access more data sources. However, even with these advances, enterprises still need agent building frameworks and platforms that sit on top of LLMs to handle the complexity of real business automation.

"Today like you know if you just work with an LLM uh that's not enough for you to actually build an agent and automate a business you know business process with it um you know you do need you know some more work you know that needs to happen uh on top of it" - Arvind Jain"

Glean addresses this through their agent builder UI, where users can converse in natural language to build agents collaboratively, while Glean handles the behind-the-scenes orchestration and API-based integrations to enterprise data sources.

Timestamp: [20:47-22:14]

Michael raises an important strategic question about the evolving competitive landscape. Major model providers like OpenAI and Anthropic are increasingly offering enterprise features and ways to connect different parts of businesses. This creates competition from multiple directions: application-layer companies building on top of models, and now model providers themselves moving into enterprise solutions.

The challenge for Glean is maintaining competitive advantage when facing this multi-front competition, especially as the large AI companies have significantly more resources and can integrate capabilities at the foundational model level.

This represents a classic platform competition dynamic where the underlying infrastructure providers begin competing with companies in the application layer, potentially commoditizing the services that sit on top of their technology.

Timestamp: [22:14-22:54]

Arvind makes a bold prediction about the fundamental transformation of work over the next 5-10 years. He asserts that the majority of work performed today will be automated by AI assistants and agents, fundamentally changing the nature of human work rather than eliminating it entirely.

"A lot of things are going to change about work in in 5 to 10 years from now majority of the work that we do today we won't do anymore uh it's going to be done by you know AI you know assistants and uh agents" - Arvind Jain"

Human work will evolve to become more creative and thinking-oriented, while base-level tasks like data manipulation, research, and analysis will be handled by AI models. This extends beyond individual tasks to entire organizational structures and business processes, which will also be automated through AI systems.

This transformation represents a shift from humans doing routine work to humans focusing on higher-level creative and strategic thinking while AI handles operational execution.

Timestamp: [22:54-24:15]

Rather than viewing competition as a threat, Arvind reframes the landscape as having enormous white space and opportunity. He argues that the volume of work that needs to be accomplished far exceeds the capacity of companies currently addressing these challenges. Even if all players work extremely hard, they're collectively addressing only about 1% of what needs to happen in the next year.

"The body of work that needs to get done um versus the companies that are actually doing that work there's a big delta between that you know the white space is huge" - Arvind Jain"

This perspective suggests the market is so large and the transformation so comprehensive that there's room for many successful companies rather than a zero-sum competitive environment. The challenge isn't too many competitors chasing too few opportunities, but rather the opposite—massive demand with insufficient supply of solutions.

Timestamp: [24:15-24:47]

Arvind articulates Glean's strategic approach to the competitive landscape: maintain partnerships with all major players rather than viewing them as competitors. This includes continuing partnerships with language model companies and cloud hyperscalers, fully leveraging their innovation rather than trying to replicate it.

The reasoning is practical: these large players won't be able to cover more than 10% of what needs to be accomplished, leaving substantial opportunity for specialized companies like Glean. As the foundational technology improves, Glean benefits by building on top of these advances.

"We will continue to partner with all of these you know players we'll continue to partner with the language model companies with the cloud hyperscalers and we'll you know fully leverage you know all the innovation that they are uh making" - Arvind Jain"

This symbiotic relationship means that as the core technology companies build more capabilities, Glean can rise along with them while focusing on areas they haven't addressed, rather than competing directly with their massive R&D investments.

Timestamp: [24:47-25:35]

From a practical customer perspective, Arvind addresses the challenge enterprises face when being pitched by hundreds of different AI vendors. Glean's unique positioning comes from explicitly not being an LLM company, which creates a clear differentiation in the marketplace.

When enterprises think about their AI strategy, they need to consider various components of the technology stack. Glean's message to customers is clear about their role and focus, helping enterprises understand how Glean fits into their broader AI architecture rather than competing with foundational model providers.

"We're telling our vendors that like like you know you we're not a LM company so when you actually think about your AI" - Arvind Jain"

This positioning allows Glean to complement rather than compete with the core AI infrastructure, making it easier for enterprises to understand how to integrate Glean into their existing or planned AI implementations.

Timestamp: [25:35-25:59]

• Enterprise agent building requires orchestration frameworks beyond simple LLM capabilities
• Computer use is a provider feature rather than just a model improvement, enabling broader data access
• Glean's agent builder UI allows natural language collaboration for building enterprise automation
• Major model providers are expanding into enterprise solutions, creating multi-directional competition
• The future of work will see majority of current tasks automated by AI within 5-10 years
• Human work will shift toward creative and strategic thinking while AI handles operational tasks
• The market opportunity vastly exceeds current competitive capacity—only 1% of needed work is being addressed
• Partnership strategy leverages rather than competes with foundational AI infrastructure providers
• Large AI companies can only address ~10% of enterprise AI needs, leaving substantial opportunity
• Glean's positioning as "not an LLM company" creates clear differentiation for enterprise customers
• The transformation scope includes individual tasks, business processes, and organizational structures

Timestamp: [20:47-25:59]

AI Companies:

• OpenAI - Major model provider mentioned as expanding into enterprise offerings
• Anthropic - Major model provider mentioned as expanding into enterprise offerings

Technology Concepts:

• Computer Use - LLM provider feature (not just model capability) that enables access to more data sources
• Agent Building Frameworks - Platforms that sit on top of LLMs to handle business process automation
• Cloud Hyperscalers - Major cloud computing platforms that Glean partners with

Business Concepts:

• LLM Company - Type of company focused on language model development, which Glean explicitly is not

Timestamp: [20:47-25:59]

Arvind articulates Glean's unique positioning as a horizontal AI platform rather than a vertical solution provider. When enterprises consider their AI stack, they still need to work with language model companies—either directly or through Glean. However, Glean offers something distinct: a platform connected to all enterprise information, data, and knowledge that can serve as a consolidation point for AI work.

The alternative approaches are stark. Companies using thousands of SaaS applications could end up with 10,000 different agents in the future if they pursue a functional, product-by-product approach to AI implementation. Instead of this fragmented vertical approach, enterprises can choose Glean as their horizontal AI layer and use it as a central platform to build many different agents.

"What we are doing is you know we're giving you know our customers a horizontal AI platform uh a platform that's you know connected you know to you know all of your enterprise information and data and knowledge so so you can actually use this as uh a way to consolidate your AI work" - Arvind Jain"

This represents a classic bundling versus unbundling dynamic, where Glean is taking the unbundled SaaS products and various AI models and bundling them into one coherent experience.

Timestamp: [25:59-27:26]

When Michael asks about Arvind's love for competition and how it factors into leadership, Arvind provides an engineer's perspective on competitive dynamics. As an engineer, inspiration often comes from witnessing amazing technologies developed by others, creating a drive to match and exceed those capabilities. Competition serves a crucial purpose in driving R&D engines at any enterprise.

Arvind reflects on Glean's early days when they were pioneers in applying transformers to search and AI assistant products six years ago. Being alone in the market for an extended period created unique challenges—without competitors, it was difficult to gauge whether they were moving fast enough or innovating at the right pace.

"When you're by yourself like you know it's very hard to pace it's very hard to actually figure out are you moving fast enough or not like you know because there's no comparison there's no benchmark" - Arvind Jain"

The isolation could even lead to complacency, since there was no external pressure or comparison point to drive urgency and improvement.

Timestamp: [27:26-29:24]

Competition serves multiple vital functions beyond just driving urgency. It creates essential pressure within teams by establishing clear stakes: innovate fast and stay ahead, or face obsolescence. This dynamic generates both agency and velocity that wouldn't exist in isolation.

Perhaps more importantly, competition prevents any single company from monopolizing good ideas. No company can conceive of all possible innovations independently. Competitive environments foster cross-pollination of ideas, where companies learn from each other's approaches and build upon external innovations to develop their own R&D roadmaps.

"No one company has like the monopoly and all the ideas like you know we like we can't think of all the great things like you know we learned so much from like what other people like you know come up with" - Arvind Jain"

This collaborative-competitive dynamic ultimately benefits the entire industry, as more participants working on similar challenges leads to faster overall progress and a larger total market that benefits all players.

Timestamp: [29:24-30:31]

Michael highlights Glean's remarkable recent performance, noting that the company recently announced crossing $100 million in Annual Recurring Revenue (ARR)—an impressive milestone that demonstrates significant market traction. Even more striking is that the business has tripled in size over the past year, representing extraordinary growth velocity in the enterprise software space.

This growth trajectory indicates that Glean has moved well beyond the early product-market fit challenges that Arvind described earlier in the conversation. The combination of market timing, competitive dynamics, and the ChatGPT-driven awareness of enterprise AI capabilities appears to be driving substantial demand for Glean's platform.

The $100M ARR milestone is particularly significant in the enterprise software world, as it typically indicates a company has achieved substantial scale and market validation, often serving as a key benchmark for enterprise software companies approaching potential public market readiness.

Timestamp: [30:31-30:55]

• Glean positions itself as a horizontal AI platform to consolidate enterprise AI work rather than pursuing vertical solutions
• Enterprises using thousands of SaaS applications could face managing 10,000+ agents without a consolidation strategy
• Companies can work with language model providers directly or through Glean's platform
• Glean represents a bundling play, combining unbundled SaaS products and AI models into unified experience
• Competition serves as crucial innovation driver, providing benchmarks and preventing complacency
• Being first-to-market without competitors made it difficult to gauge appropriate innovation pace
• Competition prevents monopolization of ideas and enables cross-pollination of innovations
• More industry participants ultimately benefits everyone by expanding the total market opportunity
• Glean recently crossed $100M ARR milestone, indicating significant market validation
• The business tripled in size over the past year, demonstrating extraordinary growth velocity
• Growth trajectory suggests successful transition from product-market fit challenges to scaled market adoption

Timestamp: [25:59-30:55]

Business Concepts:

• ARR (Annual Recurring Revenue) - Key SaaS metric, with Glean crossing $100 million milestone
• SaaS Applications - Software-as-a-Service products that enterprises use extensively
• Horizontal AI Platform - Glean's positioning as cross-functional AI infrastructure
• Vertical Agents - Function-specific or product-specific AI solutions
• Bundling vs Unbundling - Classic software industry dynamic that Glean is participating in

Strategic Concepts:

• AI Stack - The technology infrastructure enterprises build for AI capabilities
• R&D Engine - Research and development capabilities driven by competitive pressure

Timestamp: [25:59-30:55]

When asked about the key drivers behind Glean's remarkable growth—crossing $100M ARR and tripling revenue in the past year—Arvind attributes much of it to timing. The six years of foundational work that Glean invested before the AI boom has positioned them significantly ahead of the market when demand finally materialized.

This represents a classic case of preparation meeting opportunity. While other companies are now scrambling to build enterprise AI solutions, Glean already had sophisticated infrastructure, security frameworks, and proven integrations in place when the market suddenly became ready to adopt these technologies.

"We feel fortunate like you know that all the work you know that we did over the last uh 6 years you know has allowed us to actually create a product that is ahead of the market you know in a very significant way" - Arvind Jain"

When enterprises began actively seeking AI workplace solutions, Glean's mature offering stood out in a landscape of experimental and incomplete alternatives.

Timestamp: [31:01-31:34]

Arvind identifies a critical market dynamic that has benefited Glean: the gap between AI demonstrations and production deployment. While many companies can show impressive AI demos, enterprises struggle when attempting to deploy AI in production environments. The result is often frustration, experimentation without clear value, and uncertainty about return on investment.

This challenge stems from AI's fundamental unpredictability compared to traditional machines. Unlike conventional tools that perform tasks consistently and predictably, AI systems can provide four different answers to the same question asked four times. This inconsistency creates confusion and leads many people to abandon AI adoption after initial disappointing experiences.

"AI has changed that like you know it doesn't feel like that like you ask the same question to to it four times it's going to answer it four different ways and so sometimes you know people get very um confused you know with uh with a technology like this and they give up" - Arvind Jain"

Many potential users tried ChatGPT, experienced hallucinations or inconsistent responses, and concluded that AI was unreliable, creating widespread skepticism that enterprises must overcome.

Timestamp: [31:34-32:56]

Enterprise leaders face a fundamental conundrum: they recognize that AI will transform their businesses and they need to prepare, but the technology is challenging to implement effectively. Leaders understand they cannot afford to fall behind, yet they must prevent employee abandonment of AI tools due to complexity or frustration.

This has created a new executive priority beyond just ROI: education and cultural transformation. CEOs now focus on ensuring every employee becomes "AI-first," developing comfort and proficiency with AI tools as part of their regular work processes.

"On on the CEO's CEO's minds these days is that well look you know yes I want ROI from AI but I also want education to happen i want like you know every employee in my company to become an AI first employee" - Arvind Jain"

Glean's growth accelerates because companies view it as the most accessible tool for widespread employee adoption, helping workers become familiar with AI integration in their daily tasks while delivering immediate practical value.

Timestamp: [32:56-33:57]

When Michael asks how Glean drives AI-first culture within their own organization, Arvind reveals a surprising insight: even at a generative AI native company, employees don't automatically integrate AI into their daily work. Humans are creatures of habit who continue established workflows even when better alternatives exist.

This created an embarrassing situation where Glean was evangelizing AI adoption to customers while struggling to achieve full AI integration within their own team. While employees embraced Glean's product, they weren't exploring other AI tools or showing curiosity about the broader AI landscape.

"It was sort of frankly embarrassing that like you know we are going out there to our market and talking evangelizing AI when we can't get our own employees to actually fully embrace it" - Arvind Jain"

The solution required top-down initiatives to force behavioral change, recognizing that even motivated, knowledgeable employees need structural encouragement to break existing habits.

Timestamp: [33:57-34:55]

To address internal AI adoption challenges, Arvind implemented a simple but effective mandate: every executive leader must identify and implement one AI use case per quarter. The size or complexity of the use case doesn't matter—the goal is consistent experimentation and integration.

This approach acknowledges that busy employees won't naturally explore AI applications without explicit encouragement. By making AI experimentation a formal expectation for leadership, it creates accountability and demonstrates organizational commitment to AI adoption.

"We said that look you know you know I want every executive leader you know to actually come up with one use case um in a quarter and it could be a small use case I don't care how big or small it is but one use case you know that you'll actually start to use AI uh and LLMs" - Arvind Jain"

Glean recommends this same approach to their customers, recognizing that successful AI transformation requires deliberate top-down initiatives to overcome natural human resistance to workflow changes.

Timestamp: [34:55-35:34]

When asked about differences between founding Rubric and Glean, Arvind emphasizes that the experiences couldn't be more different due to the nature of the products, domains, and his personal role evolution. This suggests significant learning and growth between the two ventures, with each presenting unique challenges and opportunities.

The comparison promises insights into how entrepreneurial approaches evolve with experience, how different market sectors require different strategies, and how founder roles change between first and second startup experiences.

Timestamp: [35:34-35:59]

• Six years of pre-boom development positioned Glean significantly ahead when AI demand materialized
• Major gap exists between impressive AI demos and successful production deployment in enterprises
• AI's unpredictability (different responses to same questions) creates user confusion and abandonment
• Enterprise leaders face conundrum: must adopt AI to stay competitive but technology is difficult to implement
• CEO priorities now include AI education and creating "AI-first employees," not just ROI
• Glean grows rapidly because it's viewed as most accessible tool for company-wide AI adoption
• Even AI-native companies struggle with internal AI adoption due to human habit patterns
• Top-down initiatives are necessary to force AI experimentation and workflow changes
• "One use case per quarter" rule provides practical framework for systematic AI integration
• Successful AI transformation requires deliberate organizational structure, not just access to tools
• Second startup experience differs completely from first due to product nature, domain, and founder evolution

Timestamp: [31:01-35:59]

AI Tools:

• ChatGPT - Referenced as example of AI tool that many people tried and abandoned due to inconsistent responses
• LLMs - Language models that executives are expected to experiment with quarterly

Business Concepts:

• ROI (Return on Investment) - Traditional business metric that CEOs still want from AI implementations
• AI-First Employee - New concept of workers who are comfortable integrating AI into their regular workflows
• Dog Fooding - Practice of using your own product internally (mentioned by Michael)
• OKRs - Objectives and Key Results framework (mentioned by Michael as potential culture driver)

Companies:

• Rubric - Arvind's previous company, referenced for comparison with Glean founding experience

Timestamp: [31:01-35:59]

When asked about key learnings from Rubric that he applied to Glean, Arvind emphasizes that building successful startups fundamentally comes down to assembling great teams. While this might seem obvious, the execution requires bringing together the best engineers, salespeople, and marketers while giving them autonomy and agency to do their best work.

During Glean's first one to two years, Arvind viewed his primary role as CEO to be a recruiter—focused entirely on building and assembling the best possible team. This philosophy, which he learned at Google and applied at Rubric, has served Glean well.

"The most important thing you could do to build a successful startup is to build a great team of engineers and it's obvious everybody would say it like but you got to bring the best people the best engineers best salespeople best marketers" - Arvind Jain"

The amazing team ultimately enabled them to build both an amazing product and an amazing business, demonstrating how foundational talent acquisition is to startup success.

Timestamp: [35:53-36:46]

Arvind highlights a fundamental difference between his two startups: market maturity. Rubric operated in an established market where customers were already purchasing data protection products. They could approach prospects who were spending a million dollars on outdated technology and offer a superior modern alternative—a classic better mousetrap scenario.

Glean faced the opposite challenge: they built a product for which no budgets existed. From day one, they knew they were creating a category rather than competing in an existing one. This meant no competition but also no established buying patterns or allocated budgets.

"We came we built a product and we knew it from day one that we're building a product for which there are no budgets um you know and and and so it's going to be a different kind of emotion like you know there's no competition but there's also no budget" - Arvind Jain"

This required a completely different approach—a long journey focused on creating budgets and educating the market about a new category of solution.

Timestamp: [36:46-37:39]

The second major difference between Rubric and Glean relates to user base and go-to-market strategy. Rubric served a specialized audience—very few people within each company used the data protection product, and Arvind often had existing relationships with these technical specialists. This allowed them to accelerate go-to-market and even sell products before they worked perfectly, because customers were willing to collaborate on product development.

Glean presents the opposite scenario: it's an end-user product that everyone in a company uses. This created a fundamental challenge because they couldn't rely on existing relationships with all potential users within an organization.

"In Rubik it like you know allowed us to actually you know accelerate you know the go to market and start to sell the product before it actually really worked properly because you know all the people that you were selling the product to were people you had relationships with" - Arvind Jain"

This required changing the entire go-to-market approach. Instead of selling early and iterating with friendly customers, they had to bring the product to a much higher level of quality and ensure it felt amazing to users before unleashing their go-to-market motion.

Timestamp: [37:39-38:40]

Michael observes that Arvind represents a technically-minded CEO type that seems increasingly common in the current AI wave, contrasting with perhaps more business or sales-oriented CEOs from previous waves. When asked about how this technical orientation has worked for Glean, Arvind provides a nuanced perspective on technical founders.

While engineers often start companies because they have the capability to build something, Arvind notes that the CEO role quickly becomes much more complicated than just product development. The role requires thinking about competition, customers, and strategy in ways that may not align with an engineering mindset focused on building products.

"I constantly have that doubt you know on like you know the uh like as the company keeps scaling like you know can I handle the complexity like you know like yes I want to learn and like hopefully like you know I can I can keep scaling" - Arvind Jain"

Despite the trend of founders staying longer in CEO roles, Arvind expresses personal doubt about his ability to handle increasing complexity as the company scales, demonstrating remarkable self-awareness and humility about the challenges of growing into the CEO role.

Timestamp: [38:45-40:51]

Despite being a successful technical founder, Arvind explicitly disagrees with the notion that engineers make better CEOs. He acknowledges that while founders are increasingly staying longer in their roles, building a business ultimately requires capabilities that extend far beyond engineering skills.

The complexity of running a company demands strategic thinking about competition, customers, and business dynamics that may not come naturally to someone with primarily an engineering background. This honest assessment demonstrates intellectual humility and suggests that successful technical CEOs must continuously develop non-technical capabilities.

"I would I would disagree with the notion of that like engineers are better CEOs i don't think so" - Arvind Jain"

This perspective is particularly valuable coming from someone who has successfully made the transition, suggesting that success comes from recognizing limitations and continuously learning rather than assuming technical skills automatically translate to business leadership.

Timestamp: [40:51-41:03]

• Building great teams is the most critical factor for startup success, requiring focus on recruiting best talent across all functions
• CEO's primary role in early years should be recruitment and team assembly
• Established markets allow faster go-to-market with existing budgets and customer buying patterns
• Category creation requires long-term investment in market education and budget development
• End-user products require higher quality standards before launch compared to specialist B2B tools
• Existing relationships enable early sales of imperfect products in B2B specialist markets
• Mass-market products must feel amazing to users before wide deployment
• Technical founders often start companies due to building capabilities but CEO role extends far beyond engineering
• Scaling complexity creates ongoing challenges for technical founders transitioning to business leadership
• Engineering skills don't automatically translate to effective CEO capabilities
• Successful technical CEOs must continuously develop non-technical business skills
• Self-awareness about limitations is crucial for founder evolution and company success

Timestamp: [35:53-41:09]

Companies:

• Rubric - Arvind's previous data protection company, used for comparison with Glean's journey
• Google - Referenced as source of team-building culture and philosophy

Business Concepts:

• Data Protection - The established market category that Rubric operated in
• Category Creation - The challenge Glean faced in building a new market category
• Go-to-Market Motion - Business strategy for bringing products to market
• R&D - Research and development investment required before scaling

Roles:

• Distinguished Engineer - Arvind's former title at Google, indicating senior technical leadership

Timestamp: [35:53-41:09]

In the rapid-fire closing segment, Michael asks Arvind about his favorite productivity hack using Glean. Arvind reveals that he now asks Glean every question he has before reaching out to colleagues directly. This serves a dual purpose: it's faster for him to get answers, and it allows him to continuously test their own product to ensure it's working properly.

But what Arvind finds most exciting is how this changes customer demonstrations. When prospects ask whether Glean can handle specific use cases, instead of just answering verbally, he puts them directly in front of the product for live demonstrations.

"I don't actually answer those questions anymore like you know myself I actually want to put them right in front of them in clean and like you know do a live demo" - Arvind Jain"

This approach showcases the product's power in real-time and creates more compelling sales experiences than traditional presentations.

Timestamp: [41:16-42:19]

When asked about underrated technology trends, Arvind provides a contrarian perspective on the current AI landscape. While the industry focuses heavily on model limitations—insufficient reasoning capabilities, the need for AGI, and waiting for the next breakthrough—Arvind believes we're missing massive opportunities with existing technology.

He argues that current models, including both frontier models and smaller open-domain models, provide amazing capabilities that we've barely begun to explore. The real opportunity lies in engineering and application development on top of existing model technology rather than waiting for the next model generation.

"There's an amazing technology at our disposal right now and we've not even used 1% of the power uh uh that you know that like you know the current models actually provide" - Arvind Jain"

This perspective suggests that the next wave of AI innovation will come from creative engineering and application development rather than just model improvements, representing significant untapped potential in the current technology stack.

Timestamp: [42:19-43:27]

When asked about a technology from science fiction that he wishes existed, Arvind describes a concept of ultimate efficiency: a system where thoughts automatically convert into answers without any friction. He envisions technology that could anticipate questions before they're even spoken and provide immediate responses.

This represents the logical endpoint of search and AI assistance—eliminating all barriers between curiosity and knowledge. Arvind admits that while he often has questions and curiosity, he sometimes lacks the energy to articulate them verbally or through typing.

"I would love to have a system where like you know thoughts like convert into answers like you know just automatically" - Arvind Jain"

Brain interface technology that enables zero-friction information access would represent the ultimate realization of Glean's mission to make enterprise knowledge instantly accessible, extending beyond current interface limitations to direct thought-to-answer systems.

Timestamp: [43:27-44:12]

Podcast Information:

• This is an encore presentation of a conversation between Lightspeed partner Michael Mignano and Arvind Jain
• Listeners are encouraged to rate, review, and subscribe to the podcast for notifications of new episodes
• Generative Now is produced by Lightspeed in partnership with Pod People

Social Media & Contact Information:

• Follow Lightspeed at @LightspeedVP on YouTube, X (Twitter), and LinkedIn
• Follow Michael Mignano at @Mignano on the same platforms
• New episodes are published weekly

Acknowledgments:

• Michael thanks Arvind for his time and insights
• Both participants express that the conversation was enjoyable and valuable

Timestamp: [44:12-44:47]

• Using your own product daily provides both productivity benefits and continuous quality assurance
• Live product demonstrations are more compelling than verbal explanations of capabilities
• Current AI models offer vastly untapped potential through engineering and application development
• Industry focus on model limitations may be missing immediate opportunities with existing technology
• Creative engineering on current models could deliver more value than waiting for next-generation breakthroughs
• Both frontier and smaller open-domain models have unexplored capabilities
• Zero-friction interfaces represent the ultimate goal for information access systems
• Brain-computer interfaces could eliminate barriers between curiosity and knowledge
• Direct thought-to-answer systems would be the logical evolution of current search and AI assistance
• Successful product development requires balancing future vision with maximizing current technology potential

Timestamp: [41:16-44:47]

Technology Concepts:

• AGI (Artificial General Intelligence) - Advanced AI capability that the industry is working toward
• Frontier Models - The most advanced current AI models
• Open Domain Models - Smaller, more accessible AI models available to developers
• Brain Interface - Neural technology that could enable direct thought-to-computer communication

Business Concepts:

• Live Demo - Real-time product demonstration approach that Arvind prefers over verbal explanations
• Zero Friction - Concept of eliminating barriers between user intent and system response

Media & Production:

• Pod People - Podcast production company partnering with Lightspeed
• Lightspeed/LightspeedVP - Venture capital firm producing the podcast
• X (Twitter) - Social media platform for following updates

Timestamp: [41:16-44:47]