 |
|
|
|||||
When Glean CEO Arvind Jain thinks about the future of work, he doesnt picture a chatbot waiting for you to type a question. He imagines a digital companion that knows your goals, your meetings, even the documents youre writing, and steps in to help before you ask. That vision led him to found Glean in 2019 after more than a decade at Google and a successful run cofounding the data management startup Rubrik. Glean began as a search tool for company data. Over time it has grown alongside AI itself, moving from simple search to synthesizing information, reasoning through tasks, and into what Jain calls agentic AI, where systems take initiative. With a valuation of $7.2 billion (and our number one Most Innovative Company in Applied AI this year), Glean is positioning itself at the center of enterprise productivityand offering a preview of how agentic AI may transform the way entire workforces operate. Fast Company spoke with Jain about Glean’s origin story, the rise of agentic AI, and how enterprises are rethinking both human and machine roles in the workplace. The conversation has been edited for length and clarity. Before Glean, you founded another startup, Rubrik, where you noticed challenges with knowledge and documents. Can you share that story? I used to work at Google as a search engineerthats my primary training. I was there for over a decade, and then in early 2014 I started a company with a few others called Rubrik, an enterprise data protection and security company. We grew fastover 2,000 people in four yearsbut eventually hit a wall in productivity. Per person, we were writing less than half as much code and selling half as much product. People were struggling, and as a founder I wanted to know why. Startups thrive on productivity, so this was an existential issue. We did surveys, which showed people saying, “Hey, look, I cannot find anything in this company. I dont know where to go for the information I need, or who to ask for help.” That mirrored my own experience; our knowledge was scattered across 300 systems, and even I often couldnt find what I needed without my EA. When complaints kept coming in, I thought we could just buy a search product to connect everything, but nothing existed. There was no enterprise search tool that unified all SaaS systems. Thats what led to the creation of Glean. Ive built search products my whole life, and I felt we could finally build one that would help every employee in every company. How did that sort of kernel lead you to agentic AI as a solution? Well, we started the company in early 2019. The word “agentic” hadnt yet come into play. AI had been a big domain in computer science for a while, but even transformersthe core technology behind todays language modelswere still niche in search teams. Initially, this technology was built at Google to make search better: to understand the worlds knowledge at a higher conceptual level and grasp peoples questions more deeply, matching content semantically rather than by keyword. We were already seeing big results with transformers in search. So when we started Glean, we knew we could use this technology to build a search product that could handle natural language, understand content conceptually, and help people find the right things with precision. AI played a big role in our first version of the product. In fact, that made us the first enterprise gen AI company in the world, though little did we know it would soon take over. Over the years, the capabilities of AI models just kept increasing. The first development was understanding knowledge conceptually. The second was the ability to write: to make their own responses and answers. That was super helpful because it let us evolve our product from looking like Google inside your work life to looking more like ChatGPT inside your company. Instead of forcing the human to read everything, we let AI read and return precise answers. The third major shift was reasoning: the ability to think more like humans. That unleashed the wave of agentic AI. Now you can come into Glean, ask it to do work for you, and its connected with all your companys data as well as the worlds knowledge. It understands tasks, goes into that agent loop, and completes complex work. As models improved, weve been able to keep layering those capabilities into our platform [Image: Glean] Youve said fundraising wasnt a necessity for Glean. Why is that? I think AI companies are having a good time. In some ways its easy to raise capital, easy to do a lot of investment and build your products. For us, were a little different in terms of company history. Were actually the first gen AI company in the enterprise, before it became super hot. So we did need to raise that first round of capital, but we didnt know wed be able to raise more unless we generated success, unless we built a profitable product. So we grew up more traditionally, building a high-margin, profitable business that pays. Over the years, weve built a good business, adding value for enterprise customers and getting paid for it. Our burn hasnt been super high. Later rounds werent motivated by a need to invest in 10 areas, but more by opportunistic timing when great investors came inbound. Ultimately, I think its a good ideaeveryone else is raising capital, and we cant be seen as behind. It gives us flexibility to invest when needed. But if I look at our balance sheet right now, we havent even tapped the last few rounds of funding we raised. Even so, you closed a Series F this year at a $7.2 billion valuation. How does that affect your ambitions? As a company we never had a goal like, “Hey, this is the year we need to go IPO or become public.” In my previous startup, we had those goals, and they were irrelevant to the actual business. We had a five-year plan to go public, but it took 10, and it didnt impact the day-to-day. When we started Glean, I never thought about that as the milestone. Instead, its about how do you keep adding value to customers and build a businessthat grows steadily every year. The latest funding round really serves to make it clear to the worlds largest enterprisesour typical customersthat were here to stay. AI is becoming more and more strategic, and every large enterprise knows they have to transform with AI. They want strong partners who will be around. You focus exclusively on enterprise customers. Whats the pitch to them over competitors? I think we should first understand what Glean is. Glean is actually two different things. One is a general-purpose conversational AI assistant. Think of it as ChatGPT, but inside a company. We leverage all the best models out there, whether its GPT, Claude, Gemini, Grok, or others, and bring them to enterprise customers. Then, when Glean is deployed, it connects with all your enterprise systems and builds a deep understanding of how your business works: Who the people are, the key projects, whos an expert on what, what knowledge is fresh, whats stale. From there, Glean Assistant becomes a go-to tool that uses both internal context and the worlds knowledge to complete tasks for anyone in the company, from HR to legal to engineering The second part is the platform. To build the assistant, we had to create deep connectivity into enterprise systems, respecting governance and permissions. You cant just train models on enterprise knowledge and make it accessible to everyone; some information is restricted. So we built a platform to power that secure experience, and many customers now use the platform directly to build their own agents and AI applications. We provide the data connectivity and retrieval technologies to make that possible. So why work with us instead of Microsoft, OpenAI, Google, or any of the other big players now competing in this space? First, were pioneerswe started six and a half years ago, more than four years before anyone else, and have built the most advanced technology stack. Second, were model agnostic, giving customers access to the best innovation across providers. Third, AI models alone know nothing about your businessyou need enterprise context, and thats what we specialize in. And lastly, large enterprises demand strict compliance and data residency requirements. Meeting those needs has been our focus from day one. This is where we really distinguish ourselves from most AI companies. [Image: Glean] What is the role of humans in this agentic AI-filled future? We tend to overthink what AI means to the world. Is AI going to take over the world? Largely I would just think of AI as yet another technologyone of the most powerful weve seen in our lifetimes, but still just a tool in our toolbox as humans. The role of humans is to leverage it, to get the best out of it, to innovate more, work faster, do things we couldnt do before. From that perspective, what we tell enterprise customers is that the role of humans, number one, is to learn and get familiar with AI and its power. Glean Assistant is a good way to make that happen because it feels like ChatGPT or Google, so theres not much to learn. Were also proactive: Glean is designed to understand you and your work life, and when we detect youre about to do something, we can say, “Hey, I can do it for you. Heres my work. You can review it and see if you like it.” Thats the first part of the human role: becoming the master of AI, learning its power, and leveraging it. The next part is changing how you work with AI. Youre still in control, but enterprises must adapt to get the best out of it. You have to digitize more of your enterprise, capture human intelligence so AI can learn from it. A lot of human judgment today isnt recorded, but it needs to be if AI is going to help effectively. At Glean, we try to help companies both capture that intelligence and leverage it for future work. Ultimately, AI will become more proactive. Today, if you use ChatGPT or Glean, you usually have to go ask it to do something, and thats where you lose leverage; people rarely change habits. Our vision is for Glean to be a personal companion at work, understanding everything about your work life: your career ambitions, annual goals, weekly priorities, daily meetings, what youre writing and reading. With that knowledge, it can proactively help before you even ask. When AI comes to you, instead of you going to AI, thats when it becomes truly democratized and everyone benefits. How do you balance massive data input with security and governance concerns? First you have to understand that information in a company is fundamentally governed. You as an individual have access to some information and not others. Any AI products or solutions have to respect that. Thats a key part of our architecture: we understand your information architecture and data governance, we know what roles people play and what information they can use. So when we do something for an individual with AI, we restrict ourselves to only the information that person is authorized to access. We focus a lot on delivering safe AI experiences. And in this future where your personal AI companion knows so much about youeven listening to a watercooler conversation with the goal of helpingyou could say theres a loss of privacy. But you should not lose privacy. In that world, there must be a level of privilege between you and your companion, like attorney-client or doctor-patient privilege, where nobody else sees what you and your assistant talk about. You have to straddle that balance, getting the best from the technology while still fully preserving privacy. When you envision these agents interacting with each other across different platforms and companies, how do you see that taking shape? Is it an open ecosystem, or a more fractured one? If you think about software, it used to be very heterogeneous and boxed; different pieces wouldnt work with each other. That was the enterprise software world two decades ago. Then SaaS came and made things more interoperable. Many SaaS products could work with each other, but it was still hard because they all had custom APIs, and you had to do a lot of work to integrate them even though the systems were technically open. With AI agents, were starting from a better position. From day one, everybody knows agents have to be interoperable. Customers are demanding it, and any smart customer today wont invest in a platform that locks them down because more and more of their business is running on these agents. With this technology, dependence on AI providers is very high, so everything you do with one vendor has to transfer easily to another. Theres great demand for interoperability, and thats what were seeing. We already have integrations with other agent platformsyou can build an agent in AWS Bedrock, Google Vertex, Flow, or Copilot Studio and still call into agents or tools built in Glean, and vice versa. So were really seeing strong interoperability. [Image: Glean] Tell me about the UX. How does agentic AI change design? One thing that has happened not just in our product, but in any SaaS product over the last 15 years, is they all get more and more complicated. You build features because customers ask for them, and now you have a product that can do a thousand things. And 99% of your users dont even know about them. Theres a constant struggle from a design perspective: You invest in building features but cant make them discoverable. So when we built our conversational assistant, we realized design has to focus on the moment: no menus, no hidden tabs. You have to understand what the user is trying to do, decide which capability makes sense, and surface it wherever their attention is. Thats been a key learning, especially with AI products whose capabilities are unbounded. Theres no way to organize them in a menu. Design is the key to getting the most out of AI, and these products will look and feel very different from the last two decades Looking ahead, whats the future for Glean? Our ambition is to be one of the most important enterprise AI companies in the world. We want to be the primary interface through which AI adds value to every person in any company. One of the limitations of AI today is that its reactive, it doesnt come to you. Glean Assistant is great at helping with tasks, but you have to go to it, and that significantly limits impact. Many people dont instinctively turn to AI; theyre creatures of habit who keep working the way they always have. So one of our key directions this year is making Glean more proactive: truly understanding individuals at a personal level in their work life, goals, and tasks, and becoming a companion thats bidirectional. Yes, you can go to your Glean companion to get things done, but it also nudges you and proactively helps with what you want to do. And thats what will truly democratize the impact of AI, because now youre not waiting for humans to be the initiators.
Category:
E-Commerce
BMW just made a subtle change to the logo on its latest car. The German automaker simplified the roundel on its new, fully electric BMW iX3 by removing the inner outlines of the logo. Most people won’t even notice. So why bother? As luxury automakers adapt to an electric future, they’re updating their branding too, and different companies have taken different approaches. Jaguar went for a big change ahead of a new product launch in 2026 with a new mark that’s lighter, rounded, and lowercase as compared to its old all-caps logo mark. Ranger Rover, meanwhile, split the difference, introducing a new secondary mark that gives the brand more flexibility. Newer EV companies often use a stenciling effect to give their brand names a sci-fi look, while General Motors’ rebranded 2021 mark also went shifted to a rounded lowercase. In broad strokes, the new logo on the iX3, the first in BMW’s next generation Neue Klasse family of electric cars, isn’t all that different from BMW’s very first logo in 1917. They’re both circular and use a blue-and-white quadrant, and though the company updates it occasionally to reflect changing design trends, the basics remain the same. The Munich, Germany-based company keeps the general idea, but updates it for the times. [Photo: BMW] In 1953, BMW swapped out a gold logo outline for white. In 1963, it changed the logo’s font from serif to sans-serif. A 1997 version used shading and gradients to create a chrome, metallic effect, and in 2020, BMW added minimalist, open version for communications only. Though BMW’s logo is believed by some to be a propeller, the circular badge shape actually comes from the logo of Rapp Motorenwerke, the aircraft engine manufacturer that became BMW. The white-and-blue quadrant pattern is actually a reference to the state colors of Bavaria. [Photo: BMW] The company says the propeller myth has become self-perpetuating, but Fred Jakobs, archive director of BMW Group Classic says it also “has acquired a certain justification.” For the new version on the roundel of the iX3, Oliver Hailer the head of BMW Design, told BMWBLOG, “We wanted to keep the heritage, but bring more precision to the logo.” If it’s not broke, don’t fix it. Just spruce it up. The details matter, but sometimes a rebrand doesn’t have to be dramatic.
Category:
E-Commerce
In South Africa, a field covered in yellow wildflowers doesnt look like an industrial site. But its a pilot for a new type of nickel mine: Instead of blasting holes in the ground to extract rocks, a biotech startup called Genomines is phytomining nickel through the use of plants that absorb the metal from the soil.The plant, a type of daisy, is known as a hyperaccumulatora species that naturally pulls metal through its roots and stores it at high concentrations in its stems and leaves. Using gene editing, Genomines made the plant three times larger and able to soak up twice as much nickel. The company, which just raised $45 million in a Series A funding round, plans to use its approach to scale up a sustainable, affordable supply of the critical metal.[Photo: Courtesy of Genomines]Its important because we need a lot of metal, especially for the energy transition in batteries in electric vehicles, says Fabien Koutchekian, cofounder and CEO of Genomines. Not only in batteries, but [nickel is] widely used in stainless steel as part of infrastructure. The problem is that with current traditional mining methods, we will not be able to produce enough.Its getting harder to find nickel ore to mine. Most of it comes from Chinese-run mines in Indonesia; high-grade reserves, used to make stainless steel, could be depleted there before the end of the decade. Lower-grade ore used in batteries might run out by midcentury.Nickel also exists in soil. But until now the concentrations have been too low to make extraction viable. The plants change the economics.The plants that we are using have the ability to concentrate the metal that they find in the soilthey concentrate it in their biomass, Koutchekian says. Weve managed to reach close to 7.6% metal within the plants.[Photo: Courtesy of Genomines]The companys pilot site in South Africa sits on land thats relatively high in nickel because of the way rocks naturally weathered in the area. That means it cant be used for farming, because other plants cant grow well. But its ideally suited for a phytomine.The crop grows within four to six months, absorbing the metal. Then it can be harvested, dried, and heated to produce battery-grade nickel oxide that can be sold and refined.[Gif: Courtesy of Genomines]Its inherently far more efficient than the existing system. Building a traditional, multibillion-dollar nickel mine involves not only a decade-plus of exploration, but another decade-plus of construction. Theyre the size of small cities, Koutchekian says. Once they’re operating, traditional mines also have to move tons of rock to extract a tiny fraction of metal.Using agriculture to get the material means that minimal infrastructure is necessary, and a system can be up and running in a year or two. Unsurprisingly, operations take far less energy than traditional mining. Since the plants also help capture CO2 as they grow, the whole process is actually carbon neutral. And instead of destroying ecosystems by blowing up habitat and creating new pollution, it helps remediate soil.Sustainability isnt the main motivator for its potential customers, Koutchekian says. Instead, they’re interested in cost: The approach saves so much energy that the product could be meaningfully less expensive than the status quo. The company expects to produce nickel oxide at around $10,000 per ton, versus an industry median of around $16,000 per ton (by the end of the decade, the average cost may rise to $19,000).With the new round of funding, led by the MIT spinout Engine Ventures, Genomines plans to use pilots to prove that its process is cost competitive. Then it will keep scaling up. The potential is large: The team has estimated that around 30 million to 40 million hectares of land worldwide contain sufficient nickel for the process. In theory, if all of that land was in use, the company says it could produce 7 to 14 times as much nickel as the traditional industry does now.
Category:
E-Commerce
All news |
||||||||||||||||||
|
||||||||||||||||||