news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

Glean hopes to beat ChatGPT in the enterprise.

Author | Mei Yi
edit| Jingyu

After more than two years of AI wars triggered by big models, all startup teams and investors are now asking: What are the real scenarios for big models? Or, more importantly, how can we get real customers and revenue?

While ordinary consumers are still excited about conversational chat assistants, AI companies have already been looking for AI landing scenarios. For example, enterprise SaaS, a $100 billion track, is crowded with AI upstarts and technology giants such as OpenAI, Anthropic, and Microsoft.

In such a crowded field, a company called Glean has won over industry giants such as Sony Electronics and Databricks with its in-house AI search products.

Recently, this company, which was founded just five years ago, received a huge amount of $200 million from Kleiner Perkins and Lightspeed Venture Partners in its Series D financing. The company's valuation soared to $2.2 billion (about 16 billion RMB), making it a well-deserved leader in the enterprise AI track.。

How did Glean do it? What’s so special about its enterprise AI search product?

01

Centralized AI Search Platform

Glean can be seen as an AI enterprise search and knowledge management platform, with main functions including: AI search, knowledge management, and work homepage.

AI search is the core function of Glean. Compared with traditional search, its advantages lie in cross-application and personalization.

Glean has created a deeply integrated workspace that provides employees with a unified interface to access all enterprise applications and services, quickly search, locate and integrate scattered information such as meeting minutes, support tickets, project archives, etc., greatly improving the speed of finding information.

For example, users can search Slack conversation information, Google document content, and information in Confluence on Glean's platform.

Not only that, users can also perform lightweight functions of connected SaaS applications on Glean, such as starting meetings directly on Glean, creating Jira documents, etc.

Glean can access 300+ SaaS products of enterprises｜Image source: Glean official website

Based on the internal integration of enterprise data, Glean not only integrates vector search and keyword search technologies that can achieve "semantic understanding", but also uses LLM to launch a generative AI search function and launch an AI assistant。

Glean's AI assistant has three main functions:：

1. AI Answers

Glean’s AI can provide customized search results based on each user’s specific needs, preferences, and access permissions. For example, when employees in different positions and geographical locations search for their OKR indicators, Glean will give each person different results. Glean also uses employee activities (such as clicking on search results) to improve the relevance of searches.

2. Expert testing

When information cannot be retrieved, Glean can also connect employees with someone who can help answer questions or complete tasks.

Employees can find internal "subject matter experts" related to the search results by clicking the "people" option. For example, when a user wants to search for "employee data retention policy", the user can use the "people" option to find the relevant person in charge of the engineering safety department.

Employees are connected to people who can help answer questions or complete tasks | Image source: Glean official website

3. Contextual recommendations

After the user selects a document or other content, he can enter the shortcut key Cmd-J / Ctrl-J to view the supplementary content and context related to this content.

For example, when a user is viewing the "Supplier Security Questionnaire" document, he or she can enter shortcut keys to view links such as "Sales Security Document" and "Summary of Common Security Issues".

Employees can view the source file | Image source: Glean official website

In addition, Glean’s AI assistant also functions as a writing and coding assistant. For example, service teams can use Glean to generate support ticket responses. Similarly, it can speed up software development tasks, such as finding programming best practices and recent code change information.

As Glean founder Arvind Jain said, “Glean is an AI platform for all company data. You can think of it as Google or ChatGPT within your company. It can answer any questions employees ask based on internal company information and cite source code.”

In addition to the search function, Glean also provides "knowledge management" and "work homepage" functions.

"Knowledge management" means that users can share and integrate related documents or links, use new short URLs to jump to interfaces, etc., so that they can navigate to commonly used resources more easily.

Use go/benefits or go/pitchdeck instead of memorizing and searching for long URLs. Image source: Glean official website

Moreover, any employee or team can add customized descriptions to documents in the application, making it easier for others to quickly understand and share the documents with others. For example, the HR team can integrate documents and links related to employee onboarding into a dedicated collection to help new employees learn about the company more quickly.

HR teams can integrate onboarding-related documents and links into a dedicated collection | Image source: Glean official website

The "Work Homepage" presents personalized functional modules on the homepage based on the different habits of users, including company announcements, employee directories, calendars, etc. Users can also pin important items to the top of the search results.

It can be seen that Glean can connect, protect, index and understand customers' enterprise data on a centralized AI platform, which not only greatly improves the efficiency of information retrieval, but also makes information and knowledge within the enterprise easier to manage and utilize.

In this regard, Glean founder Arvind Jain said, "With these powerful updates, Glean takes enterprise search to a new level. We are providing a complementary system that enables enterprise users to stay connected not only to company knowledge, but also to each other, thereby driving progress in an increasingly complex digital work environment."

02

Let enterprises use AI with confidence

A recent Cisco survey found that more than a quarter of enterprises have banned the use of GenAI due to privacy and data security risks. In the poll, companies said they were concerned that GenAI tools would leak their IP or potentially disclose other sensitive information to the public or competitors.

In this regard, Arvind Jain, CEO of Glean, said, "Enterprise leaders have seen the power of ChatGPT in the consumer field and are eager to use its potential to significantly improve productivity and performance in the workplace. But enterprise data is very complex and there are many obstacles to overcome, including the illusion of a universal big model and the risk of data leakage. If it is not deployed properly, it is possible to make costly mistakes. It needs to be built on the right search foundation to truly realize its value."

He added, “Glean is the only company that has figured out how to combine enterprise knowledge with the reasoning power of LLM to deliver accurate, safe conversational AI experiences at work.”

So, how does Glean achieve "accuracy" and "security"?

The answer is to help companies use their own data to train their own generative AI models.

The foundation of this model is the "trusted knowledge model" that Glean has spent four years developing. This model not only understands the search content, but also the context, the relationship between people, the company's internal language, and privacy and security parameters, so it can meet the accuracy, security, and reference capabilities that match the needs of the enterprise.

The Trusted Knowledge Model is built around three pillars：

The "Trusted Knowledge Model" revolves around three pillars |｜Image source: Glean official website

1、Company Knowledge and Context:

Glean connects to all applications of the client company through more than 100 connectors, crawls data sources, and then indexes all metadata. By comprehensively combing and deeply understanding the company's internal language, internal relationships, content activities, etc., it builds a unique enterprise "knowledge graph" as a "search index" for each customer, thereby ensuring the personalization and relevance of search answers.

The "knowledge graph" not only weighs the direct connection between each piece of information, but also weighs countless other signals and relationships, such as being able to identify subtle differences, which makes the search engine's knowledge more complete and enables generative AI to continuously learn and improve, improving search relevance.

Comprehensively sort out and deeply understand the "knowledge graph" of internal enterprise information | Image source: Glean official website

2、Permissions and data management:

Glean's data security measures meet the highest industry standards and the maintenance and protection of customer personal information complies with the EU General Data Protection Regulation (GDPR).

Glean uses accurate data access permissions and data encryption. For example, Glean complies with the permission rules set in the company's data sources and conducts user access reviews to implement the principle of least privilege. This means that whether it is Slack, Teams, Jira, ServiceNow, etc., employees can only get answers based on the data they are authorized to access.

At the same time, Glean limits the risk of data leakage by encrypting all data at rest using AES 256 and encrypting all data in transit using TLS 1.2+. When a user deletes a document in the underlying application, the document is also deleted from the Glean system.

Additionally, Glean provides scalable infrastructure and auditing tools to ensure sensitive data is used as intended.

3、Fully referenceable:

Glean can show the source of each message and how each response was generated. Users can clearly know the source of each message and who is responsible for it.

Therefore, when company employees make natural language-based queries, Glean’s AI assistant will understand and analyze them by leveraging generative machine learning models, and then useCompany-specific AI search engineandRetrieval-augmented Generation (RAG) technologyto retrieve the most relevant and up-to-date information, and finally input this information data into a large language model (LLM) based on the company's internalKnowledge Graph, providing employees with “accurate” and “secure” search results based on their access permissions.

Although Glean is mixing large language models to output search answers, including OpenAI's GPT-4 and Google's transformer model BERT, Glean officials said, "Given that the company's generative AI models are customized, none of the company's data will be used to train these public models and benefit external organizations, and in fact will not be retained at all."

It can be seen that Glean is equivalent to an assistant who understands both the company's situation and the preferences of each employee. The search answers are based on the "trusted knowledge model", making all information safe, private, accurate and traceable.

Not only is Glean accurate and secure, it is also very convenient to deploy and use.

Glean offers Glean Apps and Glean API, allowing enterprises to create custom AI applications anywhere they need them using natural language, or create customized AI assistants, collaborative robots, chatbots, and agents and integrate them into their workflows and be up and running within days.

Glean also provides Glean Apps and Glean API | Image source: Glean official website

In this regard, Glean founder Arvind said, "Glean's initial setup time does not exceed two hours, and it can be deployed without any engineering skills or manual fine-tuning. Whether through a web app, a new tab, a sidebar search, a native search, or a Slack command, Glean provides seamless workflow integration."

“We believe that scaling AI-generated experiences to facilitate information access and discovery is the first step to unlocking its full potential for enterprise environments. Glean is at the forefront of training models and fine-tuning LLMs in our domain to power this advancement.”

03

The crowded enterprise AI race

Started by Arvind Jain, co-founder of cloud data management company Rubrik, Glean was inspired by Jain's observation that Rubrik employees often had difficulty finding the information they needed to do their jobs, and that employees at other companies were also struggling with the same problem.

In 2019, Jain, together with former employees from Google, Microsoft, and Meta, formed a small founding team to build Glean, an AI search application for enterprise customers, in Palo Alto, the center of Silicon Valley.

Glean’s co-founder | Image source: Glean official website

As generative AI develops, Glean grows and transforms into an industry-leading GenAI solution provider, dedicated to"To provide people with the knowledge they need to change the world"。

In fact, enterprise search is not a completely innovative concept in the market. Glean has many competitors, mainly including some large companies and other startups that provide similar services, such as Microsoft SharePoint Syntex, Amazon Kendra, Google Cloud Search, Coveo, Elastic, Lucidworks, etc.

However, for the first time, Glean has successfully created a comprehensive solution that is ahead of these competitors in simplifying the deployment and operation process through its unique AI model and personalized services.

Glean's business model is pureToB's pattern, and provides two different billing methods to corporate customers.

One is a per-seat model, with the cost per user being no more than $100 per month.The other is a customized enterprise solution based on an annual contract, the total amount of which is generally between $50,000 and $100,000. It is worth mentioning that for customers who choose the latter, Glean usually provides a certain discount, making the overall solution more cost-effective and therefore more popular with customers.

In addition, Glean has established a partnership with Google Cloud Platform (GCP). Part of the fees paid by corporate customers to Glean is used to pay GCP. This fee is relatively fixed and will not change significantly with the increase in the number of users. It is similar to a basic cost, which is about US$1,000 to US$2,000 per month.

Currently, Glean has more than 70 clients, ranging from startups to Fortune 500 companies, covering a variety of industries such as technology, media, education and healthcare.

Glean’s customers｜Image source: Glean official website

Glean has naturally become a favorite of capital, with investors including Kleiner Perkins, Sequoia Capital, Lightspeed, Latitude Capital, etc.So far, Glean has successfully completed 4 rounds of financing, raising a total of 850 millionDollar, with a valuation of 2.2 billionDollar。

Glean's investors | Image source: Glean official website

Regarding the future of Glean, Arvind Jain, founder of Glean Technologies, said, "Today, the role of AI is to help you get the information you need. But soon, its function will be more powerful and will transform into AI that works with you. You will see all kinds of chatbots and systems, and we will live in a world where artificial intelligence works for us."

Glean has undoubtedly succeeded in becoming a company with a valuation of over $2 billion in just five years. But it must be said that in 2019, generative AI was not as popular as it is now. In fact, not only in Silicon Valley, but also around the world, there are already very powerful competitors, including companies that focus on both consumer and B-side, such as OpenAI; there are also commercial software giants such as Microsoft, not to mention a bunch of AI startups behind them.

As a leader, how Glean can maintain its advantages and expand its influence in the enterprise market is a question that people are looking forward to.

*Header image source: Cloudera

This article is an original article from Geek Park. For reprinting, please contact Geek Jun on WeChat: geekparkGO

Geek Question

What do you think of Glean’s business model?

Musk: WeChat is really amazing. How about we just copy WeChat?

Like and followGeek Park Video Account，